#  @xenovacom Xenova Xenova posts on X about model, demo, code, check the most. They currently have [---------] followers and [---] posts still getting attention that total [-----] engagements in the last [--] hours. ### Engagements: [-----] [#](/creator/twitter::4560264219/interactions)  - [--] Week [-------] +5,489% - [--] Month [-------] +3,423% - [--] Months [-------] +214% - [--] Year [-------] -77% ### Mentions: [--] [#](/creator/twitter::4560264219/posts_active)  ### Followers: [---------] [#](/creator/twitter::4560264219/followers)  - [--] Month [------] +1.80% - [--] Months [------] +12% - [--] Year [------] +22% ### CreatorRank: [-------] [#](/creator/twitter::4560264219/influencer_rank)  ### Social Influence **Social category influence** [technology brands](/list/technology-brands) [stocks](/list/stocks) [social networks](/list/social-networks) [events](/list/events) [automotive brands](/list/automotive-brands) [finance](/list/finance) **Social topic influence** [model](/topic/model), [demo](/topic/demo), [code](/topic/code), [check](/topic/check), [ai](/topic/ai), [image](/topic/image), [inference](/topic/inference) #148, [in the](/topic/in-the), [realtime](/topic/realtime), [a new](/topic/a-new) **Top accounts mentioned or mentioned by** [@huggingface](/creator/undefined) [@alibabaqwen](/creator/undefined) [@huggingfacetransformers](/creator/undefined) [@karpathy](/creator/undefined) [@vercels](/creator/undefined) [@sinclanich](/creator/undefined) [@briaais](/creator/undefined) [@simonw](/creator/undefined) [@karpathys](/creator/undefined) [@next](/creator/undefined) [@ldenoue](/creator/undefined) [@fleetwood](/creator/undefined) [@vikhyatk](/creator/undefined) [@alonsosilva](/creator/undefined) [@evilpingwin](/creator/undefined) [@deepseekai](/creator/undefined) [@jinaai](/creator/undefined) [@bytedanceoss](/creator/undefined) [@clementdelangue](/creator/undefined) [@tobi](/creator/undefined) **Top assets mentioned** [Alphabet Inc Class A (GOOGL)](/topic/$googl) [Microsoft Corp. (MSFT)](/topic/microsoft) [IBM (IBM)](/topic/ibm) ### Top Social Posts Top posts by engagements in the last [--] hours "@ldenoue Good news I found this issue. It was a single line call to Math.max in the whisper logits processor 🙄 For some reason the "recommended" way to find the maximum element in a list is super error-prone. Pushing a fix now" [X Link](https://x.com/xenovacom/status/1663196116815425537) 2023-05-29T14:50Z 10.4K followers, [--] engagements "Transformers.js just hit [----] stars on GitHub 🤯 The #WebML community is growing so fast and I'm proud to be a part of it 🤗 If you ever plan on adding in-browser machine-learning functionality to your website or web-app check out the project:" [X Link](https://x.com/xenovacom/status/1677306173156990978) 2023-07-08T03:02Z [----] followers, 30.5K engagements "AI code completion running 100% locally inside your browser thanks to @BigCodeProject's StarCoder models and 🤗 Transformers.js We also got their new 1B model running at [--] tokens per second in Node.js (CPU). 🚀 Check out the demo" [X Link](https://x.com/xenovacom/status/1684896502991269889) 2023-07-30T16:38Z [----] followers, 21.2K engagements "Two annoying things about OpenAI's tokenizer playground: (1) it's capped at 50k characters and (2) it doesn't support GPT-4 or GPT-3.5. So I built my own version w/ Transformers.js It can tokenize the entire "Great Gatsby" (269k chars) in 200ms 🚀 https://huggingface.co/spaces/Xenova/the-tokenizer-playground https://huggingface.co/spaces/Xenova/the-tokenizer-playground" [X Link](https://x.com/anyuser/status/1687553025815003136) 2023-08-04T19:56Z 15.1K followers, 48.3K engagements "Since everything runs 100% locally in your browser you can get real-time feedback as you type 🔥 PS: Check out 🤗 Transformers.js if you haven't already:" [X Link](https://x.com/xenovacom/status/1687553028608389120) 2023-08-04T19:56Z [----] followers, [----] engagements "@simonpfish I love this - so creative 🔥 We need more emojis though especially 🤗 Also Id be keen to turn this into a web app w/ transformers.js So let me know when you release the code" [X Link](https://x.com/xenovacom/status/1687663218250977280) 2023-08-05T03:14Z 10.7K followers, [---] engagements "Here's a sneak peek of my "Chat with YouTube" browser extension made with @Vercel's AI SDK 🔥 It uses Llama-v2 (7B) deployed with @HuggingFace inference endpoints. 🚀 Source code and tutorial coming soon 🤗" [X Link](https://x.com/xenovacom/status/1689677763614052352) 2023-08-10T16:39Z [----] followers, [----] engagements "Transformers.js v2.5.2 now supports audio classification w/ MMS and wav2vec2 meaning you can for example perform language identification for over [----] languages 🤯🔥 Get started in just [--] lines of code 👇 Full release notes:" [X Link](https://x.com/xenovacom/status/1691204182626234368) 2023-08-14T21:44Z [----] followers, 24.1K engagements "Introducing Chat with YouTube an AI-powered browser extension that lets you chat with YouTube videos 🚀 This project shows how easy it is to build conversational browser extensions using 🤗 Inference Endpoints and @Vercel's AI SDK. + it's open source https://github.com/xenova/chat-with-youtube https://github.com/xenova/chat-with-youtube" [X Link](https://x.com/anyuser/status/1691578817012187537) 2023-08-15T22:33Z 15.1K followers, 52.7K engagements "@smjain All you really need to do is structure your model repo like ours (with all onnx files in an onnx subfolder). You can also use our conversation script to help with this (and it also does quantization):" [X Link](https://x.com/xenovacom/status/1699094574998704305) 2023-09-05T16:18Z [----] followers, [---] engagements "WOW 🤯 a GPT token probability visualizer now running client-side with Transformers.js 🤗 Check it out 👇" [X Link](https://x.com/xenovacom/status/1699192012115771800) 2023-09-05T22:45Z [----] followers, [----] engagements "We just released Transformers.js v2.6.0 New features: - [--] new architectures: BLOOM MPT BeiT CamemBERT CodeLlama GPT-J mBART ResNet WavLM and more 😍 - Over [---] newly-converted models on the Hub 🚀 - Huge model size reductions (up to -40%) 👇" [X Link](https://x.com/xenovacom/status/1700171184954819047) 2023-09-08T15:36Z [----] followers, 39.8K engagements "WebGPU-accelerated automatic speech recognition running completely in your browser 😍 What an amazing job @fleetwood___ 🚀 Go check it out 👇" [X Link](https://x.com/xenovacom/status/1701597370138640611) 2023-09-12T14:03Z [----] followers, [----] engagements "WOW 🤯 An in-browser version of ChatGPT (or HF Chat) built with 🤗 Transformers.js Yes that's right everything runs 100% locally in your browser meaning no need for a server Check it out 🔗 https://huggingface.co/spaces/mithril-security/blind_chat 🚀Just released #BlindChat: an open-source & privacy-first ChatGPT alternative 🔒BlindChat combines @huggingface transformers.js (from the great @xenovacom) with #ChatUI for a private and fully in-browser experience Try it on HF: https://t.co/REr4LA0c9r https://huggingface.co/spaces/mithril-security/blind_chat 🚀Just released #BlindChat: an" [X Link](https://x.com/anyuser/status/1704910846986682581) 2023-09-21T17:29Z 15.1K followers, 105.6K engagements "When do you *really* need to use a vector database 🤔 To try answer that question I recreated my semantic image search application to run 100% in-browser with Transformers.js (no server). After loading the model and database it only takes 50ms to compute text embeddings and perform similarity search across 25k images. No vector DB just plain old JavaScript. I think @karpathy was onto something. 👀 What do you think 🔗 Demo site: Source code: https://github.com/xenova/transformers.js/tree/main/examples/semantic-image-search-client" [X Link](https://x.com/anyuser/status/1705385934072742015) 2023-09-23T00:57Z 15.1K followers, 186.8K engagements "Transformers.js v2.6.2 now supports Document Question Answering meaning you can easily extract information from images. directly in your browser (no server needed) 🤯 We also added new models like Donut LongT5 and Blenderbot 🥳 I can't wait to see what you build 🤗" [X Link](https://x.com/xenovacom/status/1707039687842357390) 2023-09-27T14:29Z [----] followers, 14.6K engagements "Which tasks/models should we add next 🤔 Let us know in the comments 👇 #WebML Full release notes:" [X Link](https://x.com/xenovacom/status/1707039691378168278) 2023-09-27T14:29Z [----] followers, [---] engagements "@ldenoue Could you open up a GitHub issue Will make debugging and helping a lot easier 🤗" [X Link](https://x.com/xenovacom/status/1709292410600136835) 2023-10-03T19:40Z [----] followers, [--] engagements "Woah 🤯 A new 20M parameter embeddings model that gives similar performance to OpenAI's text-embedding-ada-002 but is much smaller+faster 🔥 I don't understand why 95% of developers still use closed-source embeddings models. 😅 + it's compatible with 🤗 Transformers.js New embeddings model gte-tiny is published Distilled from gte-small offering slightly-worse performance with half the layers. (Alternatively same size but better performance compared to all-MiniLM-L6-v2.) ONNX models also available. Check it out (link below) https://t.co/ogARt355Ne New embeddings model gte-tiny is published" [X Link](https://x.com/anyuser/status/1710347897793810586) 2023-10-06T17:34Z 15.1K followers, 102.8K engagements "We just released 🤗 Transformers.js v2.7.0 which adds supports for 🗣 Text to Speech w/ speecht5. This means you can now synthesize human speech directly in your browser. no server required 🔥 Check out the demo 👇 https://huggingface.co/spaces/Xenova/text-to-speech-client https://huggingface.co/spaces/Xenova/text-to-speech-client" [X Link](https://x.com/anyuser/status/1716486565701021875) 2023-10-23T16:07Z 15.1K followers, 66.4K engagements "A new open-source embeddings model with 8K context length that matches the performance of text-embedding-ada-002 🤯 This is a game changer 🔥 And now it's compatible with 🤗 Transformers.js meaning you can generate embeddings in your browser Node.js or even Deno Introducing jina-embeddings-v2 the world's first open-source model boasting an 8K context length. Matching the prowess of OpenAI's proprietary models now accessible on @huggingface signaling a significant milestone in the landscape of text embeddings. https://t.co/E4fRgTLsR4 Introducing jina-embeddings-v2 the world's first open-source" [X Link](https://x.com/anyuser/status/1717904546481992094) 2023-10-27T14:02Z 15.1K followers, 36.2K engagements "Introducing Distil-Whisper Web: 49% smaller 4.2x faster Whisper directly in your browser 🚀 Here is a side-by-side comparison with OpenAI's original version 🤯" [X Link](https://x.com/anyuser/status/1720460890560975103) 2023-11-03T15:20Z 15.1K followers, 273.8K engagements "Yes that's right. the new Distil-Whisper models from @huggingface are fully compatible with Transformers.js 🤗 This means you can generate high-quality transcripts directly in JavaScript: in-browser Node or even Deno 🤯🚀 Get started in just [--] lines of code:" [X Link](https://x.com/xenovacom/status/1720916876010635364) 2023-11-04T21:32Z [----] followers, 41.4K engagements "This is an absolute game changer 🤯 @threejs but for Gaussian Splatting 🔥 I cant wait to see what the web-dev community builds with this 🤗 cc @mrdoob" [X Link](https://x.com/xenovacom/status/1721575768898080872) 2023-11-06T17:10Z [----] followers, 26.2K engagements "We just released 🤗 Transformers.js v2.8.0 which adds a ton of new features including: 🖼 Super-resolution and image restoration w/ Swin2SR ✍ Optical character recognition w/ TrOCR 💬 Text-generation w/ Mistral and Falcon (1B params) More details in 🧵👇" [X Link](https://x.com/anyuser/status/1722661501180256311) 2023-11-09T17:04Z 15.1K followers, 46.6K engagements "Transformers.js just hit [--] million total requests on @jsDelivr with 52% of them coming in the past [--] days alone 🤯 We have a ton of exciting updates coming soon so stay tuned I'm excited to show you what's next. 🚀🔥" [X Link](https://x.com/xenovacom/status/1723765940611879265) 2023-11-12T18:13Z [----] followers, 39.3K engagements "🤗 Transformers.js just hit 5K stars on GitHub 🌟 Thank you to everyone in the community for your support and contributions. this is why open source is the best 🔥 PS: Stay tuned for some exciting updates coming soon 🚀" [X Link](https://x.com/xenovacom/status/1730947839654568302) 2023-12-02T13:51Z [----] followers, 10.7K engagements "Say goodbye to silent performance issues when prompting LLMs Today we released 🤗 Transformers.js v2.12 which adds support for chat templating 💬 This means you can generate LLM inputs for almost any model on the @huggingface Hub directly in your browser w/ JavaScript 🤯" [X Link](https://x.com/xenovacom/status/1736906358497202268) 2023-12-19T00:28Z [----] followers, [----] engagements "🚨 We're kicking off [----] with several improvements for Transformers.js developers: - Conditional typing of pipelines based on task. - Inline documentation + code snippets. - Pipeline-specific call parameters and return types. What should we add next Let us know 🤗👇" [X Link](https://x.com/anyuser/status/1742561498478645306) 2024-01-03T15:00Z [--] followers, [----] engagements "Image-to-LaTeX in [--] lines of JavaScript code with 🤗 Transformers.js This is made possible thanks to @vikparuchuri's amazing texify2 model which we converted to ONNX so it can run in the browser 🔥 💡 Project idea: browser extension to convert PDFs/screenshots to LaTeX" [X Link](https://x.com/xenovacom/status/1746499201184182650) 2024-01-14T11:47Z [----] followers, 12.6K engagements "@VikParuchuri Yes I noticed that so I made a custom ONNX export config with Optimum 🔥 Validation tests passed with 1e-5 atol on the logits (randomised inputs)" [X Link](https://x.com/xenovacom/status/1746565405320114646) 2024-01-14T16:10Z [----] followers, [--] engagements "Introducing the 🛝 Jinja Playground: Design LLM chat templates directly in your browser with instant feedback. Built with @huggingface/jinja a minimalistic JavaScript implementation of the Jinja templating engine specifically designed for parsing + rendering chat templates" [X Link](https://x.com/xenovacom/status/1749756308641697946) 2024-01-23T11:29Z [----] followers, 18.5K engagements "🚨 Hugging Chat Assistants are out 🚨 Just like OpenAI's GPTs you can now create your own personal assistant in Hugging Chat 🤯 To test it out I built a "Prisoner Interrogation" game where you must try to extract a secret password from a prisoner. Can you do it 🤔" [X Link](https://x.com/xenovacom/status/1753436385367785604) 2024-02-02T15:13Z [----] followers, 26.2K engagements "Local background removal Figma plugin built with 🤗 Transformers.js and BRIA AI's RMBG-v1.4 model This shows what an amazing opportunity it is for JavaScript developers to build powerful AI applications without worrying about API/server costs Great work @enzostvs 🔥" [X Link](https://x.com/xenovacom/status/1759585012230320600) 2024-02-19T14:25Z [----] followers, 14.7K engagements "@RemiCadene @Tesla Welcome to the team 🤗🤗🤗" [X Link](https://x.com/xenovacom/status/1765757361057091907) 2024-03-07T15:12Z [----] followers, [---] engagements "We just updated our in-browser Background Removal demo to use WebGPU and it's now 50x faster 🤯 [--] seconds down to 180ms ⚡ Powered by @bria_ai_'s RMBG-v1.4 model and 🤗 Transformers.js . and yes the video is in real time 🤯" [X Link](https://x.com/anyuser/status/1766573868024524967) 2024-03-09T21:16Z 15.1K followers, 76.6K engagements "@anotherjesse This is so cool 🔥 You should try out the WebGPU version (I've tested and it should be compatible with CLIP). You can try it out here: (see first comment for installation and running instructions)" [X Link](https://x.com/xenovacom/status/1769107950142906575) 2024-03-16T21:06Z [----] followers, [---] engagements "Grok-1 is finally out 😍 But while everyone was focused on the weights I decided to take a look at the tokenizer. I also added it to the Tokenizer Playground Structurally it looks quite similar to the Llama [--] tokenizer (BPE w/ byte-fallback) with a vocabulary size of [--] = [------] (much larger than Llama's 32k). add_prefix_space=True so "hello world" becomes [-----] [----] // "hello" "world" This is the same as Llama but different to GPT and Gemma. In addition to the "PAD" "BOS" "EOS" and "UNK" special tokens there are "separator" and "mask" tokens as well as [--] control tokens of the form "controlX"" [X Link](https://x.com/anyuser/status/1769546092574630069) 2024-03-18T02:07Z 15.1K followers, 30.5K engagements "New features in 🤗 Transformers.js v2.16.1: 🔥 New models: APISR for Anime Super-Resolution and EfficientNet for image classification 🖼 New pipeline: Image Feature Extraction 💬 Improved chat templating support: C4AI Command-R tool and RAG prompt generation See 🧵 for more info" [X Link](https://x.com/anyuser/status/1770507294087807082) 2024-03-20T17:46Z 15.1K followers, 26.1K engagements "Snowflake just released Arctic Embed a collection of open-source text embedding models optimized for retrieval accuracy and efficiency ❄ 📄 Apache [---] license 🌐 Great for in-browser use w/ 🤗 Transformers.js (22 [---] M params) ⚡ WebGPU-accelerated (120x faster than WASM)" [X Link](https://x.com/xenovacom/status/1780900393788018996) 2024-04-18T10:05Z [----] followers, [----] engagements "Meta's Llama [--] is here with a brand new tokenizer 🦙 I've added it to the Tokenizer Playground so you can experiment with it in your browser: For those interested here are the key differences over Llama 2: [--]. 4x larger vocabulary (32K - 128K). This means improved token efficiency (up to 15% fewer tokens compared to Llama 2) as well as stronger multilingualism. However it comes at the cost of much larger input and output embedding matrices which account for a sizeable portion of the parameter count increase of the small model going from 7B in Llama [--] to 8B in Llama [--]. [--]. Llama [--] uses a" [X Link](https://x.com/xenovacom/status/1780998994920923542) 2024-04-18T16:37Z [----] followers, 12.8K engagements "Meta's Segment Anything Model (SAM) can now run in your browser w/ WebGPU (+ fp16) meaning up to 8x faster image encoding (10s 1.25s) 🤯⚡ Video is not sped up Everything runs 100% locally thanks to 🤗 Transformers.js and onnxruntime-web 🔗 Demo: https://hf.co/spaces/Xenova/segment-anything-webgpu https://hf.co/spaces/Xenova/segment-anything-webgpu" [X Link](https://x.com/anyuser/status/1781846540685897965) 2024-04-21T00:44Z 15.1K followers, 122.4K engagements "Introducing Phi-3 WebGPU a private and powerful AI chatbot that runs locally in your browser powered by 🤗 Transformers.js and onnxruntime-web 🔒 On-device inference: no data sent to a server ⚡ WebGPU-accelerated ( [--] t/s) 📥 Model downloaded once and cached Try it out 👇" [X Link](https://x.com/anyuser/status/1788177160227660079) 2024-05-08T12:00Z 15.1K followers, 119K engagements "Phi-3 running at [--] tokens per second 100% locally in your browser 🤯⚡ What speed do you get Try it out yourself 👇 https://huggingface.co/spaces/Xenova/experimental-phi3-webgpu https://huggingface.co/spaces/Xenova/experimental-phi3-webgpu The always brilliant @xenovacom released an experimental WebGPU powered demo of phi-3 running directly in the browser. My consumer but well-specced workstation got [--] tokens per second The age of usable on-device inference is rapidly approaching https://t.co/1ve1FGAoBd https://huggingface.co/spaces/Xenova/experimental-phi3-webgpu" [X Link](https://x.com/xenovacom/status/1788664184487432679) 2024-05-09T20:15Z [----] followers, 19.2K engagements "@magikarp_tokens This is because Llama-3 switched from sentencepiece BPE to tiktoken BPE which actually introduced many illegal tokens (those that cant be reached using the merges according to the standard BPE algorithm). v0.19.1 fixed this here: More info here: https://github.com/huggingface/tokenizers/pull/1493 So what's the deal with tiktoken vs. sentencepiece Surely the BPE algorithm is universal Unfortunately it's not that simple. (image 1) The difference can be traced back to this section of code in tiktoken: https://t.co/ujKC3LZg94 At first it looks like an optimization to" [X Link](https://x.com/xenovacom/status/1789611642646327658) 2024-05-12T11:00Z [----] followers, [---] engagements "The GPT-4o tokenizer is so much better than GPT-4 for non-English languages 🤯 Here's a demo across [--] different languages: [----] [---] (-52%) tokens for the exact same text (1660 characters) 🔥 GPT-4o (left) vs. GPT-4 (right)" [X Link](https://x.com/anyuser/status/1790125428000145685) 2024-05-13T21:02Z 15.1K followers, 36.9K engagements "WOW I can't believe that Transformers.js was mentioned on stage at Google I/O 🤯 This is a dream come true 🤗" [X Link](https://x.com/anyuser/status/1790512129918800366) 2024-05-14T22:38Z 15.1K followers, 29.2K engagements "Moondream your favorite tiny vision language model by @vikhyatk can now run directly in the browser on WebGPU 🤯 Powered of course by Transformers.js and ONNX Runtime Web 🤗 Local inference means no data leaves your device This is huge for privacy Try it out yourself 👇" [X Link](https://x.com/anyuser/status/1791436796498174047) 2024-05-17T11:53Z 15.1K followers, 38.5K engagements "You can now use 🤗 Transformers.js with Google Visual Blocks a visual programming framework that lets you create machine learning pipelines in a no-code graph editor 🛠 Rapid workflow prototyping 💡 Intuitive drag-and-drop interface ⚡ WebGPU-accelerated in-browser inference" [X Link](https://x.com/anyuser/status/1792570966272336074) 2024-05-20T14:59Z 15.1K followers, 22.7K engagements "Mistral-7B-Instruct-v0.3 is out and now supports function calling It has an updated tokenizer so we added it to The Tokenizer Playground. 🔥 If you're learning about tokenization needing to count tokens or just want to play around with it then check out our web demo 🤗" [X Link](https://x.com/xenovacom/status/1793387482739830841) 2024-05-22T21:04Z [----] followers, [---] engagements "Transformers.js is being added to Firefox [---] 🤯 Thats right fully private on-device AI directly in your browser 🔥 The first use-case theyre exploring is automatic alt-text generation for images. A huge win for accessibility What other features would you like to see 👇 Firefox [---] is bringing a game-changing feature: automatic alt-text generation for images using a fully private on-device AI model 🙌🏾 Initially available in the built-in PDF editor our aim is to extend this to general browsing for screen reader users. https://t.co/019T4CbcW6 Firefox [---] is bringing a game-changing feature:" [X Link](https://x.com/anyuser/status/1797285648572821840) 2024-06-02T15:14Z 15.1K followers, 44.7K engagements "@jpohhhh @realmrfakename @cartesia_ai Should be possible. 👀 Others have been able to create working ONNX versions of popular diffusion models like stable diffusion so I don't see what's stopping you from getting StyleTTS working :) Keep me updated on your progress" [X Link](https://x.com/xenovacom/status/1798125360220897501) 2024-06-04T22:51Z [----] followers, [--] engagements "Jina CLIP v1 just released: a new state-of-the-art multimodal embedding model that outperforms OpenAI CLIP in text-image retrieval 😍 We also contributed ONNX weights so it's now compatible with 🤗 Transformers.js v3 and runs with WebGPU acceleration ⚡ Try out the demo 👇" [X Link](https://x.com/anyuser/status/1798321533405847771) 2024-06-05T11:50Z 15.1K followers, 45.3K engagements "It's finally possible: real-time in-browser speech recognition with OpenAI Whisper 🤯 The model runs fully on-device using Transformers.js and ONNX Runtime Web and supports multilingual transcription across [---] different languages 🔥 Check out the demo (+ source code) 👇" [X Link](https://x.com/anyuser/status/1799110540700078422) 2024-06-07T16:05Z 15.1K followers, 266.9K engagements "@tombielecki Definitely possible - it would just require updating the initial tokens passed to the decoder. Do you have an example (in python) for this I can take a look at Feel free to open a feature request on GitHub so I can track this easier. https://github.com/xenova/transformers.js https://github.com/xenova/transformers.js" [X Link](https://x.com/xenovacom/status/1799229145391743446) 2024-06-07T23:57Z [----] followers, [----] engagements "@whitphx @maartenbreddels @alonsosilva @Alibaba_Qwen @solara_dev It certainly can 🚀 Source code: Demo: https://github.com/xenova/transformers.js/tree/v3/examples/webgpu-chat Qwen2 is here with base and instruct models released across [--] sizes: 0.5B 1.5B 7B 57B-A14B and 72B 😍 To test them out I created a WebGPU chat demo of the 0.5B version with 🤗 Transformers.js It runs 100% locally in the browser (no server required). Try it out 👇 https://t.co/PIOQ2CTzPw https://github.com/xenova/transformers.js/tree/v3/examples/webgpu-chat Qwen2 is here with base and instruct models released across 5" [X Link](https://x.com/xenovacom/status/1800136502653124714) 2024-06-10T12:02Z [----] followers, [---] engagements "Depth Anything V2 just released enabling real-time depth estimation directly in your browser with 🤗 Transformers.js and WebGPU acceleration ⚡ The smallest model is only 50MB (@ fp16) making it perfect for on-device usage 😍 Check out the demo (+ source code) 👇" [X Link](https://x.com/anyuser/status/1801672335830798654) 2024-06-14T17:45Z 15.1K followers, 42.6K engagements "Florence-2 the new vision foundation model by Microsoft can now run 100% locally in your browser on WebGPU thanks to Transformers.js 🤗🤯 It supports tasks like image captioning optical character recognition object detection and many more 😍 WOW Demo (+ source code) 👇" [X Link](https://x.com/anyuser/status/1805990110065803492) 2024-06-26T15:42Z 15.1K followers, 90.6K engagements "After my recent post about running Gemini Nano locally in your browser using Chrome's new window.ai feature many people asked how to set it up themselves. 🤔 So I wrote a blog post/tutorial explaining the process 🔥 Hope it helps 👇 https://huggingface.co/blog/Xenova/run-gemini-nano-in-your-browser https://huggingface.co/blog/Xenova/run-gemini-nano-in-your-browser" [X Link](https://x.com/xenovacom/status/1811428734097891613) 2024-07-11T15:53Z [----] followers, [----] engagements "Introducing SmolLM: a new SOTA series of 135M 360M and 1.7B models perfect for on-device deployment 🔥 We also uploaded ONNX weights for the models meaning they can run locally in your browser with 🤗 Transformers.js and WebGPU acceleration ⚡ Try it out (+ blog post) 👇" [X Link](https://x.com/anyuser/status/1813258097185448377) 2024-07-16T17:03Z 15.1K followers, 34K engagements "Mistral and NVIDIA just released Mistral NeMo a state-of-the-art 12B model with 128k context length 😍 It uses a new Tiktoken-based tokenizer which is far more efficient at compressing source code and non-English languages (3x for Korean 🤯) Tokenizer playground link 👇" [X Link](https://x.com/xenovacom/status/1813968731250274784) 2024-07-18T16:07Z [----] followers, [----] engagements "Introducing Whisper Diarization: Multilingual speech recognition with word-level timestamps and speaker segmentation running 100% locally in your browser thanks to 🤗 Transformers.js Tested on this iconic Lettermen interview w/ Grace Hopper from [----] Demo (+ source code) 👇" [X Link](https://x.com/anyuser/status/1815428390792335550) 2024-07-22T16:47Z 15.1K followers, 94.9K engagements "🤏 SmolLM Instruct v0.2 just dropped: a series of blazingly fast and remarkably powerful small language models (135M 360M and 1.7B params). Apache [---] licensed 🔥 They can even run in-browser on WebGPU with 🤗 Transformers.js at [--] tokens/second 🤯 Demo (+ source code) 👇" [X Link](https://x.com/anyuser/status/1825898023106257394) 2024-08-20T14:09Z 15.1K followers, 16.9K engagements "I can't believe this. Phi-3.5-mini (3.8B) running in-browser at [--] tokens/second on WebGPU w/ Transformers.js and ONNX Runtime Web 🤯 Since everything runs 100% locally no messages are sent to a server a huge win for privacy 🔒 Check out the demo + source code below 👇" [X Link](https://x.com/anyuser/status/1826992922509595068) 2024-08-23T14:40Z 15.1K followers, 22.3K engagements "@nmstoker Running on an RTX [----] courtesy of @evilpingwin 🤗 (who can provide more details if youd like)" [X Link](https://x.com/xenovacom/status/1826996407741337711) 2024-08-23T14:54Z 10.6K followers, [---] engagements "There has been a huge debate recently about the best approach for image background removal. Here's my attempt: - In-browser inference w/ 🤗 Transformers.js - WebGPU accelerated (fast) - Costs $0 (no image hosting or server processing) - No data leaves your device (privacy) It works. It's fast. It's super high quality. It's cheap. And it scales. Available for purchase soon https://t.co/5If4T3Inj3 It works. It's fast. It's super high quality. It's cheap. And it scales. Available for purchase soon https://t.co/5If4T3Inj3" [X Link](https://x.com/anyuser/status/1828116951186710795) 2024-08-26T17:07Z 15.1K followers, 422.3K engagements "BONUS: We're releasing the entire project under the Apache [---] license so feel free to adapt the demo and use it for commercial purposes The background removal model (MODNet) is also Apache [---] 🤗 Code: 🔗 Demo: https://huggingface.co/spaces/webml-community/remove-background-webgput=0 https://github.com/huggingface/transformers.js-examples/tree/main/remove-background-webgpu https://huggingface.co/spaces/webml-community/remove-background-webgput=0 https://github.com/huggingface/transformers.js-examples/tree/main/remove-background-webgpu" [X Link](https://x.com/xenovacom/status/1828552248915378437) 2024-08-27T21:56Z 10.6K followers, [----] engagements "@quicksave2k @fleetwood___ Indeed well work with the ONNX Runtime Web team to get this added 🤗🔥 @fleetwood___ Do you have plans to add support for them in Transformers.js I assume it means adding them to ONNX runtime web first right @fleetwood___ Do you have plans to add support for them in Transformers.js I assume it means adding them to ONNX runtime web first right" [X Link](https://x.com/xenovacom/status/1838406761688240355) 2024-09-24T02:35Z 10.6K followers, [--] engagements "Llama [---] running 100% locally in your browser on WebGPU 🦙 Up to [--] tokens per second ⚡ Powered by 🤗 Transformers.js and ONNX Runtime Web. No installation required. just visit a website Check out the demo and source code below 👇" [X Link](https://x.com/anyuser/status/1840767709317046460) 2024-09-30T14:56Z 15.1K followers, 35.2K engagements "Run OpenAI's new Whisper Turbo model 100% locally in your browser with Transformers.js ⚡ Transcribe [--] minutes of audio in [--] seconds 🤯 Demo + source code 👇" [X Link](https://x.com/anyuser/status/1841080068593905960) 2024-10-01T11:37Z 15.1K followers, 141.7K engagements "Google just released Gemma-2-JPN a fine-tuned version of Gemma [--] 2B on Japanese text. It supports the Japanese language at the same performance as English-only queries on Gemma [--] 😍 You can even run it 100% locally in your browser on WebGPU using 🤗 Transformers.js 👇 GoogleGemma-2-JPNGemma [--] 2Bfine-tuneGemma [--] : https://t.co/LgyHREpCOf GoogleGemma-2-JPNGemma [--] 2Bfine-tuneGemma [--] : https://t.co/LgyHREpCOf" [X Link](https://x.com/xenovacom/status/1841797046446981122) 2024-10-03T11:06Z 10.6K followers, [----] engagements "Google is on fire with their open source releases 🔥 Today they dropped Gemma-APS a collection of Gemma models for text-to-propositions segmentation. The models are distilled from fine-tuned Gemini Pro models applied to multi-domain synthetic data 👇 https://huggingface.co/collections/google/gemma-aps-release-66e1a42c7b9c3bd67a0ade88 https://huggingface.co/collections/google/gemma-aps-release-66e1a42c7b9c3bd67a0ade88" [X Link](https://x.com/anyuser/status/1846220407730278761) 2024-10-15T16:03Z 15.1K followers, 41.1K engagements "After more than a year of development we're excited to announce the release of 🤗 Transformers.js v3 ⚡ WebGPU support (up to 100x faster than WASM) 🔢 New quantization formats (dtypes) 🏛 [---] supported architectures in total 📂 [--] new example projects and templates 🤖 Over [----] pre-converted models 🌐 Node.js (ESM + CJS) Deno and Bun compatibility 🏡 A new home on GitHub and NPM Get started with npm i @huggingface/transformers. Learn more in the blog post below 👇" [X Link](https://x.com/anyuser/status/1848741677122654483) 2024-10-22T15:02Z 15.1K followers, 44.7K engagements "@NERDDISCO @huggingface @Microsoft @v0 Awesome 🤩🙌" [X Link](https://x.com/xenovacom/status/1851192593868001316) 2024-10-29T09:21Z 13.9K followers, [---] engagements "OmniParser the new screen parsing tool from Microsoft (and #1 trending model on @huggingface) can now run 100% locally in your browser with Transformers.js 🤯 Who's going to be the first to turn this into a browser extension 👀 Endless possibilities Demo & code below 👇" [X Link](https://x.com/anyuser/status/1852251703791288469) 2024-11-01T07:29Z 15.1K followers, 66.2K engagements "Given a screenshot of a user interface it detects interactable icons/elements on the page (w/ bounding boxes) and generates meaningful descriptions for each. It's meant to enhance the ability of a multimodal model (like GPT-4V) to generate actions (so it can interact with the page). Links to learn more: - GitHub repo: - Blog post: - Project page: https://microsoft.github.io/OmniParser/ https://www.microsoft.com/en-us/research/articles/omniparser-for-pure-vision-based-gui-agent/ https://github.com/microsoft/OmniParser https://microsoft.github.io/OmniParser/" [X Link](https://x.com/xenovacom/status/1852257146513019093) 2024-11-01T07:51Z 10.7K followers, [----] engagements "WOW 🤯 Language models are becoming smaller and more capable than ever Here's SmolLM2 running 100% locally in-browser w/ WebGPU on a 6-year-old GPU. Look at that speed ⚡😍 Powered by 🤗 Transformers.js and ONNX Runtime Web How many tokens/second do you get Let me know 👇" [X Link](https://x.com/anyuser/status/1861757975540445549) 2024-11-27T13:04Z 15.1K followers, 12.9K engagements "We just released Transformers.js v3.1 and you're not going to believe what's now possible in the browser w/ WebGPU 🤯 Let's take a look: 🔀 Janus from @deepseek_ai for unified multimodal understanding and generation (Text-to-Image and Image-Text-to-Text) 👁 Qwen2-VL from @alibaba_qwen for dynamic-resolution image understanding 🔢 JinaCLIP from @JinaAI_ for general-purpose multilingual multimodal embeddings 🌋 LLaVA-OneVision from @ByteDanceOSS for Image-Text-to-Text generation 🤸♀ ViTPose for pose estimation 📄 MGP-STR for optical character recognition (OCR) 📈 PatchTST & PatchTSMixer for" [X Link](https://x.com/anyuser/status/1862143690191290612) 2024-11-28T14:37Z 15.1K followers, 20.7K engagements "⚡ Janus WebGPU demo (+ source code): 📝 Transformers.js v3.1 release notes: https://github.com/huggingface/transformers.js/releases/tag/3.1.0 https://huggingface.co/spaces/webml-community/Janus-1.3B-WebGPU https://github.com/huggingface/transformers.js/releases/tag/3.1.0 https://huggingface.co/spaces/webml-community/Janus-1.3B-WebGPU" [X Link](https://x.com/xenovacom/status/1862144107537133984) 2024-11-28T14:38Z 10.8K followers, [----] engagements "@NielsRogge @tumble_wood @GozukaraFurkan The WebLLM team have been able to run Llama 70B on WebGPU so it should technically be possible 👀" [X Link](https://x.com/xenovacom/status/1863907741074669682) 2024-12-03T11:26Z 11.1K followers, [---] engagements "@javierluraschi Whoops I completely misread your first message. I think I assumed TTS since the original post was about TTS. For STT whisper is definitely my go-to. It comes in different sizes (ranging from 40M params to 1.5B). https://x.com/xenovacom/status/1841080068593905960 Run OpenAI's new Whisper Turbo model 100% locally in your browser with Transformers.js ⚡ Transcribe [--] minutes of audio in [--] seconds 🤯 Demo + source code 👇 https://t.co/Xj0sxwWeCX https://x.com/xenovacom/status/1841080068593905960 Run OpenAI's new Whisper Turbo model 100% locally in your browser with Transformers.js" [X Link](https://x.com/xenovacom/status/1866141270806016137) 2024-12-09T15:22Z 10.9K followers, [--] engagements "🚨 We're hiring an intern to join the WebML team @huggingface 🚨 If you're passionate about open source and want to help build JavaScript libraries that empower the next generation of web developers we'd love to hear from you 🤗 Links to apply in 🧵" [X Link](https://x.com/xenovacom/status/1866166719993274404) 2024-12-09T17:03Z 11K followers, 14.1K engagements "Europe Middle East and Africa (EMEA) Remote https://apply.workable.com/huggingface/j/656D794A0C/ https://apply.workable.com/huggingface/j/656D794A0C/" [X Link](https://x.com/xenovacom/status/1866166723847528822) 2024-12-09T17:03Z 10.9K followers, [---] engagements "Introducing Moonshine Web: real-time speech recognition running 100% locally in your browser 🚀 Faster and more accurate than Whisper 🔒 Privacy-focused (no data leaves your device) ⚡ WebGPU accelerated (w/ WASM fallback) 🔥 Powered by ONNX Runtime Web and Transformers.js" [X Link](https://x.com/anyuser/status/1869423057741230539) 2024-12-18T16:42Z 15.1K followers, 28.8K engagements "After more than [--] years we finally have a replacement for BERT 🤯 Introducing ModernBERT a family of state-of-the-art encoder-only models with 8K sequence length better downstream performance (classification retrieval) and faster processing. An absolute game-changer 👇" [X Link](https://x.com/xenovacom/status/1869787371501224262) 2024-12-19T16:50Z 11K followers, [----] engagements "@thenameless7741 @livekit Wow this looks great The model appears to be llama-based so indeed it should be compatible with Transformers.js (after some minor repo restructuring). Let me make a PR https://huggingface.co/livekit/turn-detector https://huggingface.co/livekit/turn-detector" [X Link](https://x.com/xenovacom/status/1870577326099980473) 2024-12-21T21:09Z 11K followers, [--] engagements "Is this the future of AI browser agents 👀 WebGPU-accelerated reasoning LLMs are now supported in Transformers.js 🤯 Here's MiniThinky-v2 (1B) running 100% locally in the browser at [--] tps (no API calls) I can't wait to see what you build with it Demo + source code in 🧵👇" [X Link](https://x.com/anyuser/status/1877689001475260488) 2025-01-10T12:08Z 15.1K followers, 83.5K engagements "For the AI builders out there: imagine what could be achieved with a browser extension that (1) uses a powerful reasoning LLM (2) runs 100% locally & privately and (3) can directly access/manipulate the DOM 👀 🧑💻 Code: 🔗 Demo: https://huggingface.co/spaces/webml-community/llama-3.2-reasoning-webgpu https://github.com/huggingface/transformers.js-examples/tree/main/llama-3.2-reasoning-webgpu https://huggingface.co/spaces/webml-community/llama-3.2-reasoning-webgpu https://github.com/huggingface/transformers.js-examples/tree/main/llama-3.2-reasoning-webgpu" [X Link](https://x.com/xenovacom/status/1877689004092539125) 2025-01-10T12:08Z 11.3K followers, [----] engagements "@ClementDelangue @tobi Literally working on this right now 😂" [X Link](https://x.com/xenovacom/status/1878858431052103681) 2025-01-13T17:35Z 11.4K followers, [----] engagements "@ClementDelangue @tobi I've already uploaded Transformers.js-compatible checkpoints to the HF hub: and now I'm working on a JS implementation of the phonemizer needed to get it running in the browser: https://github.com/xenova/phonemizer.js https://huggingface.co/onnx-community/Kokoro-82M-ONNX https://github.com/xenova/phonemizer.js https://huggingface.co/onnx-community/Kokoro-82M-ONNX" [X Link](https://x.com/xenovacom/status/1878858986784763949) 2025-01-13T17:37Z 11.4K followers, [----] engagements "Introducing Kokoro.js a new JavaScript library for running Kokoro TTS an [--] million parameter text-to-speech model 100% locally in the browser w/ WASM. Powered by 🤗 Transformers.js. WebGPU support coming soon 👉 npm i kokoro-js 👈 Link to demo (+ sample code) in 🧵" [X Link](https://x.com/anyuser/status/1879904441132056633) 2025-01-16T14:52Z 15.1K followers, 42.6K engagements "Introducing SmolVLM 256M (& 500M): The worlds smallest multimodal model. Designed for efficiency and perfect for on-device applications 🔥 Its so small it can even run 100% locally in your browser on WebGPU 🤏 Powered by Transformers.js ⚡ Try it out yourself 👇" [X Link](https://x.com/anyuser/status/1882435994160447587) 2025-01-23T14:31Z 15.1K followers, 25.9K engagements "Reasoning models like o3 and o4-mini are advancing faster than ever but imagine what will be possible when they can run locally in your browser 🤯 Well with 🤗 Transformers.js you can do just that Here's Zyphra's new ZR1 model running at over [---] tokens/second on WebGPU ⚡" [X Link](https://x.com/xenovacom/status/1912579362735739113) 2025-04-16T18:50Z 12.7K followers, [----] engagements "I seriously cannot believe this is a 0.6B LLM 🤯 @Alibaba_Qwen just released Qwen3 a series of hybrid reasoning models that allow you to control how much "thinking" the model does for a given task. They can even run locally in your browser on WebGPU with 🤗 Transformers.js" [X Link](https://x.com/anyuser/status/1917032653825454382) 2025-04-29T01:46Z 15.1K followers, 76.4K engagements "Let's go 😍 @Alibaba_Qwen just released Qwen3-Embedding a new series of embedding models: 🏆 SOTA performance on MMTEB MTEB and MTEB-Code 📏 Three different sizes (0.6B / 4B / 8B) 🌍 Multilingual (119 languages) 💻 Can run in-browser w/ Transformers.js (+ WebGPU acceleration)" [X Link](https://x.com/anyuser/status/1931082176788906006) 2025-06-06T20:14Z 15.1K followers, 50.9K engagements "Browsers are the perfect place for hybrid AI inference combining the power of cloud AI with the versatility and privacy of local models. ⚡ Dia already has great WebGPU support so I'm looking forward to seeing more on-device models being integrated directly into the browser itself. especially smaller task-specific models for TTS and background removal. Great work @joshm and team I'm excited to see what's next for @diabrowser 🔥 We are floored by the reaction to @diabrowser. Thank you everyone 🥹 Theres so much more to do is an understatement. But I wanted to pause to extend heartfelt gratitude" [X Link](https://x.com/xenovacom/status/1935092938922758481) 2025-06-17T21:51Z 13.2K followers, [----] engagements "You can now run our new SmolLM3-3B model directly in your browser on WebGPU ⚡ 🏆 SoTA for its size 🧠 Dual mode reasoning 🌍 Multilingual (English French Spanish German Italian Portuguese) 🤗 Fully open source (model data code & recipes) Try it out yourself 👇 We just released the best 3B model 100% open-source open dataset architecture details exact data mixtures and full training recipe including pre-training mid-training post-training and synthetic data generation for everyone to train their own. Let's go open-source AI We just released the best 3B model 100% open-source open dataset" [X Link](https://x.com/xenovacom/status/1943801959045378180) 2025-07-11T22:37Z 13.2K followers, 10.2K engagements "@nic_o_martin @huggingface Exciting times ahead 🚀😍" [X Link](https://x.com/xenovacom/status/1947324611688837209) 2025-07-21T15:55Z 13.2K followers, [---] engagements "Did you know you can vibe code Transformers.js web-apps with @_akhaliq's Anycoder 🤯 Here's an AI image describer powered by Moondream running locally in-browser with WebGPU acceleration and deployed to Hugging Face spaces in one click ⚡ Try it out yourself 👇 Kimi K2 + @GroqInc vibe coding in anycoder build a transformers.js app for Xenova/moondream2 model card info and example included in prompt app was deployed successfully on HF in one click https://t.co/hNW2Diuihx Kimi K2 + @GroqInc vibe coding in anycoder build a transformers.js app for Xenova/moondream2 model card info and example" [X Link](https://x.com/xenovacom/status/1947345458688413714) 2025-07-21T17:18Z 13.3K followers, 15.5K engagements "A community member trained a tiny Llama model (23M parameters) on [--] million high-quality @lichess games then deployed it to run entirely in-browser with 🤗 Transformers.js Super cool 🔥 It has an estimated ELO of [----]. can you beat it 👀 (runs on both mobile and desktop)" [X Link](https://x.com/xenovacom/status/1947694341683613731) 2025-07-22T16:24Z 13.9K followers, [----] engagements "@alonsosilva @lichess Pretty much yes The highest-scoring legal move is selected on each turn. (in this case the creator just performs a single forward pass for every turn so a logits processor isn't used directly. but it's the same principle)" [X Link](https://x.com/xenovacom/status/1947798732302381409) 2025-07-22T23:19Z 13.9K followers, [---] engagements "Transformers.js just crossed 250K weekly downloads on NPM 🚀🤯 Huge thanks to the community for your continued support it's been incredible to see what you've built so far 🤗 Big updates coming soon. 👀" [X Link](https://x.com/xenovacom/status/1949979220123267527) 2025-07-28T23:44Z 13.3K followers, [----] engagements "We've made significant optimizations to our ONNX export process enabling 2x faster inference for BERT models on WebGPU ⚡ 📉 Before: decomposed operations (slow) 📈 After: unified Multi-Head Attention node (fast) If you're using Transformers.js or ONNX Runtime try it out 🔥" [X Link](https://x.com/xenovacom/status/1951375648083022313) 2025-08-01T20:13Z 13.4K followers, [----] engagements "HUGE: OpenAI just released GPT OSS on Hugging Face 🤯 Here's what you need to know: [--]. Two models: gpt-oss-20b and gpt-oss-120b designed for powerful reasoning agentic tasks and more [--]. Mixture-of-experts (MoE) architecture: 21B and 117B total parameters with 3.6B and 5.1B active parameters respectively. [--]. 4-bit quantization scheme (MXFP4) only applied on the MoE weights allowing the 20GB to fit in a single 16GB GPU and the 120B to fit in a single 80GB GPU. [--]. Text-only reasoning models with chain-of-thought and adjustable reasoning effort levels. [--]. Instruction-tuned with support for tool" [X Link](https://x.com/xenovacom/status/1952779231013871750) 2025-08-05T17:10Z 13.9K followers, [----] engagements "Did you know you can try the new gpt-oss models (120b & 20b) directly from their model cards 🤯 Look at that speed 🚀 Check it out 👇" [X Link](https://x.com/xenovacom/status/1952797159897604467) 2025-08-05T18:21Z 13.3K followers, [----] engagements "gpt-oss-20b: gpt-oss-120b: https://huggingface.co/openai/gpt-oss-120b https://huggingface.co/openai/gpt-oss-20b https://huggingface.co/openai/gpt-oss-120b https://huggingface.co/openai/gpt-oss-20b" [X Link](https://x.com/xenovacom/status/1952797492363264195) 2025-08-05T18:22Z 13.3K followers, [---] engagements "gpt-oss-120b (reasoning: high) absolutely crushes the "bouncing ball in a spinning hexagon" test. Try it out in the gpt-oss playground: http://gpt-oss.com http://gpt-oss.com" [X Link](https://x.com/xenovacom/status/1952805669687459949) 2025-08-05T18:55Z 13.4K followers, 23.8K engagements "@HCSolakoglu Official demo from OpenAI powered by Hugging Face's Inference Providers 🤗 https://huggingface.co/docs/inference-providers https://huggingface.co/docs/inference-providers" [X Link](https://x.com/xenovacom/status/1952809190797234396) 2025-08-05T19:09Z 13.3K followers, [---] engagements "@simonw @lmstudio @ollama @OpenRouterAI @CerebrasSystems Was trying this as you put out your tweet Prompt: "Generate an SVG of a pelican riding a bicycle" Adding "detailed" to the prompt produces a bicycle with spokes. pretty cool" [X Link](https://x.com/xenovacom/status/1952831927955697969) 2025-08-05T20:39Z 13.3K followers, [---] engagements "@simonw @lmstudio @ollama @OpenRouterAI @CerebrasSystems (this was the 120b running with reasoning: high)" [X Link](https://x.com/xenovacom/status/1952832381322133643) 2025-08-05T20:41Z 13.3K followers, [---] engagements "There's a new tiny TTS model in town: Kitten TTS 🐱 With just 15M parameters (25 MB) it delivers impressive quality for its size and can even run in real time without a GPU. So I created a web demo for it: featuring text normalization chunking and real-time playback. 🤗 Introducing Kitten TTS a SOTA tiny text-to-speech model - Just 15M parameters - Runs without a GPU - Model size less than [--] MB - Multiple high-quality voices - Ultra-fast - even runs on low-end edge devices Github and HF links below https://t.co/9T3u1M0WGo Introducing Kitten TTS a SOTA tiny text-to-speech model - Just 15M" [X Link](https://x.com/anyuser/status/1953227206240481705) 2025-08-06T22:50Z 15.1K followers, 14.1K engagements "Google just released their smallest Gemma model ever: Gemma [--] 270M 🤯 🤏 Highly compact & efficient 🤖 Strong instruction-following capabilities 🔧 Perfect candidate for fine-tuning It's so tiny that it can even run 100% locally in your browser with Transformers.js 🤗" [X Link](https://x.com/anyuser/status/1956026993545203822) 2025-08-14T16:15Z 15.1K followers, 35.4K engagements "@simonw Haha thanks 😁 We dont currently support in-browser fine-tuning but youre not the first (and certainly not the last) to ask about it 👀 Maybe one day 😅" [X Link](https://x.com/xenovacom/status/1956047384049234320) 2025-08-14T17:36Z 13.4K followers, [---] engagements "Okay this is insane. WebGPU-accelerated semantic video tracking powered by DINOv3 and Transformers.js 🤯 This will revolutionize AI-powered video editors. which can now run 100% locally in your browser no server inference required (costs $0) 😍 Who's building this" [X Link](https://x.com/anyuser/status/1958893388645753181) 2025-08-22T14:05Z 15.1K followers, 43.7K engagements "@LiquidAI_ @KarnikShreyas It always amazes me to see what the community is capable of building with this stuff. 🤗 Great work @KarnikShreyas I tested it by analyzing the config of Grok-2:" [X Link](https://x.com/xenovacom/status/1960747990185849328) 2025-08-27T16:55Z 13.8K followers, [---] engagements "NEW: Google releases EmbeddingGemma a state-of-the-art multilingual embedding model perfect for on-device use cases At only 308M params the model can run 100% locally in your browser 🤯 Explore your documents in an interactive 3D universe with our demo: "The Semantic Galaxy"" [X Link](https://x.com/anyuser/status/1963638442664829051) 2025-09-04T16:21Z 15.1K followers, 23.9K engagements "IBM just released Granite [---] their latest series of small language models These models excel at agentic workflows (tool calling) document analysis RAG and more. 🚀 The "Micro" (3.4B) model can even run 100% locally in your browser on WebGPU powered by 🤗 Transformers.js" [X Link](https://x.com/anyuser/status/1973784183492485277) 2025-10-02T16:16Z 15.1K followers, 84.7K engagements "🗂 Model collection: 🔗 WebGPU demo + source code: https://huggingface.co/spaces/ibm-granite/Granite-4.0-WebGPU https://huggingface.co/collections/ibm-granite/granite-40-language-models-6811a18b820ef362d9e5a82c https://huggingface.co/spaces/ibm-granite/Granite-4.0-WebGPU https://huggingface.co/collections/ibm-granite/granite-40-language-models-6811a18b820ef362d9e5a82c" [X Link](https://x.com/xenovacom/status/1973784185589662067) 2025-10-02T16:16Z 14.1K followers, [----] engagements "🔗 Demo + source code: https://huggingface.co/spaces/ibm-granite/granite-docling-258M-WebGPU https://huggingface.co/spaces/ibm-granite/granite-docling-258M-WebGPU" [X Link](https://x.com/xenovacom/status/1975648963865747784) 2025-10-07T19:46Z 14.2K followers, [----] engagements "Chrome's new window.ai feature is going to change the web forever 🤯 It allows you to run Gemini Nano a powerful 3.25B parameter LLM 100% locally in your browser We've also added experimental support to 🤗 Transformers.js making it super easy to use 😍 Check it out 👇" [X Link](https://x.com/anyuser/status/1810356703826977183) 2024-07-08T16:54Z 15.1K followers, 583.8K engagements "First project of 2025: Vision Transformer Explorer I built a web app to interactively explore the self-attention maps produced by ViTs. This explains what the model is focusing on when making predictions and provides insights into its inner workings 🤯 Try it out yourself 👇" [X Link](https://x.com/xenovacom/status/1874471468480000349) 2025-01-01T15:03Z 12.5K followers, [----] engagements ""DeepSeek-R1-Distill-Qwen-1.5B outperforms GPT-4o and Claude-3.5-Sonnet on math benchmarks with 28.9% on AIME and 83.9% on MATH." We now have open-source reasoning models that outperform GPT-4o and run 100% locally in your browser on WebGPU 🤯 [----] is off to a wild start 🔥" [X Link](https://x.com/anyuser/status/1881805730497565124) 2025-01-21T20:47Z 15.1K followers, 174.6K engagements "DeepSeek is on FIRE 🔥 They just released Janus Pro: a multimodal LLM capable of visual understanding and image generation 🤯 The 1B model can even run in your browser on WebGPU powered by 🤗 Transformers.js This is the easiest way to run it locally: just visit a website" [X Link](https://x.com/anyuser/status/1884023431051821313) 2025-01-27T23:39Z 15.1K followers, 244.6K engagements "We did it. Kokoro TTS (v1.0) can now run 100% locally in your browser w/ WebGPU acceleration. Real-time text-to-speech without a server. ⚡ Generate [--] seconds of speech in [--] second for $0. What will you build 🔥 If someone wants a gig here is what I'd like: - convert to WebGPU - implement sentence splitting - register some nifty url with just a upload text or past from clipboard button (like https://t.co/b8lbYALU5u or whatever) - that instantly starts an audio player - save prior If someone wants a gig here is what I'd like: - convert to WebGPU - implement sentence splitting - register some" [X Link](https://x.com/anyuser/status/1887883185494565006) 2025-02-07T15:16Z 15.1K followers, 69.8K engagements "Introducing Kokoro Web: ML-powered speech synthesis directly in your browser 🚀 It's 100% free and supports unlimited generations. Choose from [--] unique voices offering natural and expressive speech for any application. Now w/ streaming & WebGPU acceleration Check it out 👇" [X Link](https://x.com/anyuser/status/1890454815525736567) 2025-02-14T17:35Z 15.1K followers, 37.4K engagements "NEW: Real-time conversational AI models can now run 100% locally in your browser 🤯 🔐 Privacy by design (no data leaves your device) 💰 Completely free. forever 📦 Zero installation required just visit a website ⚡ Blazingly-fast WebGPU-accelerated inference Try it out 👇" [X Link](https://x.com/anyuser/status/1930331292408639489) 2025-06-04T18:30Z 15.1K followers, 14.2K engagements "Introducing Granite Docling WebGPU 🐣 State-of-the-art document parsing 100% locally in your browser 🤯 🔐 No data sent to a server (private & secure) 💰 Completely free. forever 🔂 Docling ecosystem enables conversion to HTML Markdown JSON and more Try out the demo 👇" [X Link](https://x.com/anyuser/status/1975648512009511166) 2025-10-07T19:44Z 15.1K followers, 57.1K engagements "BOOM 💥 Today I added WebGPU support for @karpathy's nanochat models meaning they can run 100% locally in your browser (no server) The d32 version runs at over [--] tps on my M4 Max 🚀 Pretty wild that you can now deploy AI applications using just a single index.html file 😅 Excited to release new repo: nanochat (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining nanochat is a minimal from scratch full-stack training/inference pipeline of a simple ChatGPT clone in a single https://t.co/LLhbLCoZFt Excited to release new repo:" [X Link](https://x.com/anyuser/status/1980399601812389908) 2025-10-20T22:23Z 15.1K followers, 96.4K engagements "IBM just released Granite-4.0 Nano their smallest LLMs ever (300M & 1B) 😍 The models demonstrate remarkable instruction following and tool calling capabilities and can even run locally in-browser This means they can interact with websites and call browser APIs for you 🤯" [X Link](https://x.com/anyuser/status/1983218720366326002) 2025-10-28T17:06Z 15.1K followers, 33.1K engagements "I can't wait to see what the community builds with it 🤗 Links: 📄 Blog post: 🚀 Demo (+ source code): https://huggingface.co/spaces/ibm-granite/Granite-4.0-Nano-WebGPU https://huggingface.co/blog/ibm-granite/granite-4-nano https://huggingface.co/spaces/ibm-granite/Granite-4.0-Nano-WebGPU https://huggingface.co/blog/ibm-granite/granite-4-nano" [X Link](https://x.com/xenovacom/status/1983218722006249759) 2025-10-28T17:06Z 14.4K followers, [----] engagements "Meta Superintelligence Labs just released SAM 3: Segment Anything with Concepts. A unified model that detects segments and tracks objects in images or videos using concept prompts (text or images) 🤯 Model weights (+ demo) now out on 🤗 Hugging Face" [X Link](https://x.com/xenovacom/status/1991179734126883110) 2025-11-19T16:20Z 14.5K followers, [----] engagements "SAM [--] can even run 100% locally in your browser on WebGPU enabling real-time image segmentation ⚡ Original model: ONNX model: WebGPU demo: https://huggingface.co/spaces/webml-community/SAM3-Tracker-WebGPU https://huggingface.co/onnx-community/sam3-tracker-ONNX https://huggingface.co/facebook/sam3 https://huggingface.co/spaces/webml-community/SAM3-Tracker-WebGPU https://huggingface.co/onnx-community/sam3-tracker-ONNX https://huggingface.co/facebook/sam3" [X Link](https://x.com/xenovacom/status/1991188671022637243) 2025-11-19T16:55Z 14.5K followers, [----] engagements "Brilliant 🤩 This is a project Ive been wanting to see for so long Congrats on getting in done 🤗 If you want to go a step further: aligning the API exactly with PyTorch so that students can learn/practice how to build neural networks directly in-browser would be an invaluable teaching resource. Behind the scenes of course would need to be some transpilation *magic* to add async/awaits (to make JS happy) 😁 Excited to see the progression of this project" [X Link](https://x.com/xenovacom/status/1991882990335541686) 2025-11-21T14:54Z 14.4K followers, [---] engagements "NEW: @MistralAI releases Mistral [--] a family of multimodal models including three start-of-the-art dense models (3B 8B and 14B) and Mistral Large [--] (675B 41B active). All Apache [---] 🤗 Surprisingly the 3B is small enough to run 100% locally in your browser on WebGPU 🤯 https://twitter.com/i/web/status/1995879338583945635 https://twitter.com/i/web/status/1995879338583945635" [X Link](https://x.com/xenovacom/status/1995879338583945635) 2025-12-02T15:34Z 14.9K followers, 224.9K engagements "Link to collections: - Mistral Large [--] (675B 41B active): - Dense models (3B 8B and 14B): Try out the demo yourself 👇 https://huggingface.co/spaces/mistralai/Ministral_3B_WebGPU https://huggingface.co/collections/mistralai/ministral-3 https://huggingface.co/collections/mistralai/mistral-large-3 https://huggingface.co/spaces/mistralai/Ministral_3B_WebGPU https://huggingface.co/collections/mistralai/ministral-3 https://huggingface.co/collections/mistralai/mistral-large-3" [X Link](https://x.com/xenovacom/status/1995879340320649250) 2025-12-02T15:34Z 14.9K followers, [----] engagements "Ultra-fast expressive voice cloning is now possible with Chatterbox Turbo 🤯 Use paralinguistic tags like chuckle or gasp to enhance the realism of your generations and clone voices with just [--] seconds of reference audio. Now available on Hugging Face 🤗" [X Link](https://x.com/xenovacom/status/2000657494553821338) 2025-12-15T20:01Z 14.9K followers, 10.9K engagements "NEW: Google releases FunctionGemma a lightweight (270M) open foundation model built for creating specialized function calling models 🤯 To test it out I built a small game: use natural language to solve fun physics simulation puzzles running 100% locally in your browser 🕹" [X Link](https://x.com/xenovacom/status/2001703932968452365) 2025-12-18T17:19Z 14.9K followers, 57.1K engagements "microgpt.js: the most atomic way to train and inference a GPT in pure dependency-free JavaScript. This repository covers the complete algorithm. Everything else is just efficiency. Links: - Demo: - GitHub: https://github.com/xenova/microgpt.js https://huggingface.co/spaces/webml-community/microgpt.js https://github.com/xenova/microgpt.js https://huggingface.co/spaces/webml-community/microgpt.js" [X Link](https://x.com/xenovacom/status/2022438211931267290) 2026-02-13T22:30Z 15.1K followers, [----] engagements "Introducing Voxtral WebGPU: State-of-the-art audio transcription directly in your browser 🤯 🗣 Transcribe videos meeting notes songs and more 🔐 Runs on-device meaning no data is sent to a server 🌎 Multilingual (8 languages) 🤗 Completely free (forever) & open source https://twitter.com/i/web/status/1948404315934077354 https://twitter.com/i/web/status/1948404315934077354" [X Link](https://x.com/anyuser/status/1948404315934077354) 2025-07-24T15:26Z 15.1K followers, 20.6K engagements "🫡 Good luck to your Mac It should work provided you have enough VRAM. 🤞 https://twitter.com/i/web/status/2022358009616204151 https://twitter.com/i/web/status/2022358009616204151" [X Link](https://x.com/xenovacom/status/2022358009616204151) 2026-02-13T17:11Z 15.1K followers, [---] engagements "After nearly a year of development 🤗 Transformers.js v4 Preview is finally out on npm npm i @huggingface/transformers@next Build WebGPU-accelerated AI applications that run everywhere: browsers Node.js Bun Deno Electron and more. See what's new in our blog post 👇" [X Link](https://x.com/anyuser/status/2020920634944155800) 2026-02-09T18:00Z 15.1K followers, 25.4K engagements "Behold. GPT-OSS (20B) running 100% locally in your browser on WebGPU. This shouldn't be possible but with Transformers.js v4 and ONNX Runtime Web it is A new class of AI apps is emerging. Zero-install infinite distribution. Simply visit a website and run models locally" [X Link](https://x.com/xenovacom/status/2022352305656627280) 2026-02-13T16:48Z 15.1K followers, 304.4K engagements "Inspired by @karpathy's microgpt I built microgpt.js: a JavaScript port that runs entirely in your browser It's an exact numerical implementation so the randomness and outputs match bit-for-bit Try it out yourself and train your own GPT by simply opening a webpage 👇 New art project. Train and inference GPT in [---] lines of pure dependency-free Python. This is the *full* algorithmic content of what is needed. Everything else is just for efficiency. I cannot simplify this any further. https://t.co/HmiRrQugnP New art project. Train and inference GPT in [---] lines of pure dependency-free Python." [X Link](https://x.com/xenovacom/status/2022438209498808467) 2026-02-13T22:30Z 15.1K followers, 48.4K engagements "I get over [--] tokens/second on my M4 Max. what about you Try it out yourself and let me know 👇 https://huggingface.co/spaces/webml-community/GPT-OSS-WebGPU https://huggingface.co/spaces/webml-community/GPT-OSS-WebGPU" [X Link](https://x.com/xenovacom/status/2022352307648864603) 2026-02-13T16:48Z 15.1K followers, [----] engagements "Learn more about what else is now possible with 🤗 Transformers.js v4 https://x.com/xenovacom/status/2020920634944155800 After nearly a year of development 🤗 Transformers.js v4 Preview is finally out on npm npm i @huggingface/transformers@next Build WebGPU-accelerated AI applications that run everywhere: browsers Node.js Bun Deno Electron and more. See what's new in our blog post 👇 https://t.co/dX4m5GMkbx https://x.com/xenovacom/status/2020920634944155800 After nearly a year of development 🤗 Transformers.js v4 Preview is finally out on npm npm i @huggingface/transformers@next Build" [X Link](https://x.com/xenovacom/status/2022466397084012589) 2026-02-14T00:22Z 15.1K followers, [----] engagements "If you're not following what's happening with small specialized language models you're not paying close enough attention. 4B models can be trained to solve Olympiad-level math problems just as well as models 30x larger (GPT-OSS-120B). They can even run in-browser on WebGPU" [X Link](https://x.com/xenovacom/status/2023423565698334742) 2026-02-16T15:45Z 15.1K followers, [----] engagements "QED-Nano WebGPU: Frontier-level math AI running 100% locally in your browser. Powered by Transformers.js v4 - Blog post: - Model: - Demo (& source code): https://huggingface.co/spaces/webml-community/QED-Nano-WebGPU https://huggingface.co/lm-provers/QED-Nano https://huggingface.co/spaces/lm-provers/qed-nano-blogpost https://huggingface.co/spaces/webml-community/QED-Nano-WebGPU https://huggingface.co/lm-provers/QED-Nano https://huggingface.co/spaces/lm-provers/qed-nano-blogpost" [X Link](https://x.com/xenovacom/status/2023423568361730364) 2026-02-16T15:45Z 15.1K followers, [---] engagements "After nearly a year of development 🤗 Transformers.js v4 Preview is finally out on npm npm i @huggingface/transformers@next Build WebGPU-accelerated AI applications that run everywhere: browsers Node.js Bun Deno Electron and more. See what's new in our blog post 👇" [X Link](https://x.com/anyuser/status/2020920634944155800) 2026-02-09T18:00Z 15.1K followers, 25.4K engagements "NEW: Apple releases FastVLM and MobileCLIP2 on Hugging Face 🤗 The models are up to 85x faster and 3.4x smaller than previous work enabling real-time VLM applications 🤯 It can even do live video captioning 100% locally in your browser (zero install). Huge for accessibility" [X Link](https://x.com/anyuser/status/1961454543503344036) 2025-08-29T15:42Z 15.1K followers, 224.7K engagements "It's finally possible: real-time in-browser speech recognition with OpenAI Whisper 🤯 The model runs fully on-device using Transformers.js and ONNX Runtime Web and supports multilingual transcription across [---] different languages 🔥 Check out the demo (+ source code) 👇" [X Link](https://x.com/anyuser/status/1799110540700078422) 2024-06-07T16:05Z 15.1K followers, 266.9K engagements "There has been a huge debate recently about the best approach for image background removal. Here's my attempt: - In-browser inference w/ 🤗 Transformers.js - WebGPU accelerated (fast) - Costs $0 (no image hosting or server processing) - No data leaves your device (privacy) It works. It's fast. It's super high quality. It's cheap. And it scales. Available for purchase soon https://t.co/5If4T3Inj3 It works. It's fast. It's super high quality. It's cheap. And it scales. Available for purchase soon https://t.co/5If4T3Inj3" [X Link](https://x.com/anyuser/status/1828116951186710795) 2024-08-26T17:07Z 15.1K followers, 422.3K engagements "It works. It's fast. It's super high quality. It's cheap. And it scales. Available for purchase soon" [X Link](https://x.com/anyuser/status/1828024167431836024) 2024-08-26T10:58Z 282.1K followers, 823.4K engagements ""DeepSeek-R1-Distill-Qwen-1.5B outperforms GPT-4o and Claude-3.5-Sonnet on math benchmarks with 28.9% on AIME and 83.9% on MATH." We now have open-source reasoning models that outperform GPT-4o and run 100% locally in your browser on WebGPU 🤯 [----] is off to a wild start 🔥" [X Link](https://x.com/anyuser/status/1881805730497565124) 2025-01-21T20:47Z 15.1K followers, 174.6K engagements "DeepSeek is on FIRE 🔥 They just released Janus Pro: a multimodal LLM capable of visual understanding and image generation 🤯 The 1B model can even run in your browser on WebGPU powered by 🤗 Transformers.js This is the easiest way to run it locally: just visit a website" [X Link](https://x.com/anyuser/status/1884023431051821313) 2025-01-27T23:39Z 15.1K followers, 244.6K engagements "Run OpenAI's new Whisper Turbo model 100% locally in your browser with Transformers.js ⚡ Transcribe [--] minutes of audio in [--] seconds 🤯 Demo + source code 👇" [X Link](https://x.com/anyuser/status/1841080068593905960) 2024-10-01T11:37Z 15.1K followers, 141.7K engagements "Introducing Whisper Diarization: Multilingual speech recognition with word-level timestamps and speaker segmentation running 100% locally in your browser thanks to 🤗 Transformers.js Tested on this iconic Lettermen interview w/ Grace Hopper from [----] Demo (+ source code) 👇" [X Link](https://x.com/anyuser/status/1815428390792335550) 2024-07-22T16:47Z 15.1K followers, 94.9K engagements "Introducing Whisper Web: ML-powered speech recognition directly in your browser 🚀 This comes with the release of Transformers.js v2.2.0 which now supports multilingual transcription and translation for over [---] different languages 🤯 Check it out: https://huggingface.co/spaces/Xenova/whisper-web https://huggingface.co/spaces/Xenova/whisper-web" [X Link](https://x.com/anyuser/status/1667210518317203457) 2023-06-09T16:42Z 15.1K followers, 1.1M engagements "I'm excited to announce that Transformers.js V3 is finally available on NPM 🔥 State-of-the-art Machine Learning for the web now with WebGPU support 🤯⚡ Install it from NPM with: @/ or via CDN (example below) 👇" [X Link](https://x.com/anyuser/status/1823026862647656515) 2024-08-12T16:00Z 15.1K followers, 89.4K engagements "Chrome's new window.ai feature is going to change the web forever 🤯 It allows you to run Gemini Nano a powerful 3.25B parameter LLM 100% locally in your browser We've also added experimental support to 🤗 Transformers.js making it super easy to use 😍 Check it out 👇" [X Link](https://x.com/anyuser/status/1810356703826977183) 2024-07-08T16:54Z 15.1K followers, 583.8K engagements "Meta's Segment Anything Model (SAM) can now run in your browser w/ WebGPU (+ fp16) meaning up to 8x faster image encoding (10s 1.25s) 🤯⚡ Video is not sped up Everything runs 100% locally thanks to 🤗 Transformers.js and onnxruntime-web 🔗 Demo: https://hf.co/spaces/Xenova/segment-anything-webgpu https://hf.co/spaces/Xenova/segment-anything-webgpu" [X Link](https://x.com/anyuser/status/1781846540685897965) 2024-04-21T00:44Z 15.1K followers, 122.4K engagements "IBM just released Granite [---] their latest series of small language models These models excel at agentic workflows (tool calling) document analysis RAG and more. 🚀 The "Micro" (3.4B) model can even run 100% locally in your browser on WebGPU powered by 🤗 Transformers.js" [X Link](https://x.com/anyuser/status/1973784183492485277) 2025-10-02T16:16Z 15.1K followers, 84.7K engagements "Introducing Whisper Timestamped: Multilingual speech recognition with word-level timestamps running 100% locally in your browser thanks to 🤗 Transformers.js This unlocks a world of possibilities for in-browser video editing 🤯 What will you build 😍 Demo (+ source code) 👇" [X Link](https://x.com/anyuser/status/1811068015229747335) 2024-07-10T16:00Z 15.1K followers, 108.2K engagements "Florence-2 the new vision foundation model by Microsoft can now run 100% locally in your browser on WebGPU thanks to Transformers.js 🤗🤯 It supports tasks like image captioning optical character recognition object detection and many more 😍 WOW Demo (+ source code) 👇" [X Link](https://x.com/anyuser/status/1805990110065803492) 2024-06-26T15:42Z 15.1K followers, 90.6K engagements "Is this the future of AI browser agents 👀 WebGPU-accelerated reasoning LLMs are now supported in Transformers.js 🤯 Here's MiniThinky-v2 (1B) running 100% locally in the browser at [--] tps (no API calls) I can't wait to see what you build with it Demo + source code in 🧵👇" [X Link](https://x.com/anyuser/status/1877689001475260488) 2025-01-10T12:08Z 15.1K followers, 83.5K engagements "Introducing Whisper WebGPU: Blazingly-fast ML-powered speech recognition directly in your browser 🚀 It supports multilingual transcription and translation across [---] languages 🤯 The model runs locally meaning no data leaves your device 😍 Check it out (+ source code) 👇" [X Link](https://x.com/anyuser/status/1799858691639796089) 2024-06-09T17:38Z 15.1K followers, 95.3K engagements "I seriously cannot believe this is a 0.6B LLM 🤯 @Alibaba_Qwen just released Qwen3 a series of hybrid reasoning models that allow you to control how much "thinking" the model does for a given task. They can even run locally in your browser on WebGPU with 🤗 Transformers.js" [X Link](https://x.com/anyuser/status/1917032653825454382) 2025-04-29T01:46Z 15.1K followers, 76.4K engagements "After more than a year of development we're excited to announce the release of 🤗 Transformers.js v3 ⚡ WebGPU support (up to 100x faster than WASM) 🔢 New quantization formats (dtypes) 🏛 [---] supported architectures in total 📂 [--] new example projects and templates 🤖 Over [----] pre-converted models 🌐 Node.js (ESM + CJS) Deno and Bun compatibility 🏡 A new home on GitHub and NPM Get started with npm i @huggingface/transformers. Learn more in the blog post below 👇" [X Link](https://x.com/anyuser/status/1848741677122654483) 2024-10-22T15:02Z 15.1K followers, 44.7K engagements "OmniParser the new screen parsing tool from Microsoft (and #1 trending model on @huggingface) can now run 100% locally in your browser with Transformers.js 🤯 Who's going to be the first to turn this into a browser extension 👀 Endless possibilities Demo & code below 👇" [X Link](https://x.com/anyuser/status/1852251703791288469) 2024-11-01T07:29Z 15.1K followers, 66.2K engagements "Introducing Phi-3 WebGPU a private and powerful AI chatbot that runs locally in your browser powered by 🤗 Transformers.js and onnxruntime-web 🔒 On-device inference: no data sent to a server ⚡ WebGPU-accelerated ( [--] t/s) 📥 Model downloaded once and cached Try it out 👇" [X Link](https://x.com/anyuser/status/1788177160227660079) 2024-05-08T12:00Z 15.1K followers, 119K engagements "Introducing Kokoro.js a new JavaScript library for running Kokoro TTS an [--] million parameter text-to-speech model 100% locally in the browser w/ WASM. Powered by 🤗 Transformers.js. WebGPU support coming soon 👉 npm i kokoro-js 👈 Link to demo (+ sample code) in 🧵" [X Link](https://x.com/anyuser/status/1879904441132056633) 2025-01-16T14:52Z 15.1K followers, 42.6K engagements "We just updated our in-browser Background Removal demo to use WebGPU and it's now 50x faster 🤯 [--] seconds down to 180ms ⚡ Powered by @bria_ai_'s RMBG-v1.4 model and 🤗 Transformers.js . and yes the video is in real time 🤯" [X Link](https://x.com/anyuser/status/1766573868024524967) 2024-03-09T21:16Z 15.1K followers, 76.6K engagements "BOOM 💥 Today I added WebGPU support for @karpathy's nanochat models meaning they can run 100% locally in your browser (no server) The d32 version runs at over [--] tps on my M4 Max 🚀 Pretty wild that you can now deploy AI applications using just a single index.html file 😅 Excited to release new repo: nanochat (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining nanochat is a minimal from scratch full-stack training/inference pipeline of a simple ChatGPT clone in a single https://t.co/LLhbLCoZFt Excited to release new repo:" [X Link](https://x.com/anyuser/status/1980399601812389908) 2025-10-20T22:23Z 15.1K followers, 96.4K engagements "Excited to release new repo: nanochat (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining nanochat is a minimal from scratch full-stack training/inference pipeline of a simple ChatGPT clone in a single dependency-minimal codebase. You boot up a cloud GPU box run a single script and in as little as [--] hours later you can talk to your own LLM in a ChatGPT-like web UI. It weighs [----] lines of imo quite clean code to: - Train the tokenizer using a new Rust implementation - Pretrain a Transformer LLM on FineWeb evaluate CORE score" [X Link](https://x.com/anyuser/status/1977755427569111362) 2025-10-13T15:16Z 1.8M followers, 5.8M engagements "Introducing Granite Docling WebGPU 🐣 State-of-the-art document parsing 100% locally in your browser 🤯 🔐 No data sent to a server (private & secure) 💰 Completely free. forever 🔂 Docling ecosystem enables conversion to HTML Markdown JSON and more Try out the demo 👇" [X Link](https://x.com/anyuser/status/1975648512009511166) 2025-10-07T19:44Z 15.1K followers, 57.1K engagements "We did it. Kokoro TTS (v1.0) can now run 100% locally in your browser w/ WebGPU acceleration. Real-time text-to-speech without a server. ⚡ Generate [--] seconds of speech in [--] second for $0. What will you build 🔥 If someone wants a gig here is what I'd like: - convert to WebGPU - implement sentence splitting - register some nifty url with just a upload text or past from clipboard button (like https://t.co/b8lbYALU5u or whatever) - that instantly starts an audio player - save prior If someone wants a gig here is what I'd like: - convert to WebGPU - implement sentence splitting - register some" [X Link](https://x.com/anyuser/status/1887883185494565006) 2025-02-07T15:16Z 15.1K followers, 69.8K engagements "If someone wants a gig here is what I'd like: - convert to WebGPU - implement sentence splitting - register some nifty url with just a upload text or past from clipboard button (like or whatever) - that instantly starts an audio player - save prior uploads and play progress in local storage - consider for downloading weights if not present no server all local essentially zero running cost public good. https://webtorrent.io/ http://read4.me https://webtorrent.io/ http://read4.me" [X Link](https://x.com/anyuser/status/1878847016744427992) 2025-01-13T16:50Z 448.9K followers, 408.9K engagements "New @karpathy video just dropped 😍🔥 After watching if you want to learn more about how different models (e.g. GPT4 Llama T5 BERT) tokenize text check out "The Tokenizer Playground": a web-app I built a few months ago with 🤗 Transformers.js 🔗 https://huggingface.co/spaces/Xenova/the-tokenizer-playground New (2h13m 😅) lecture: "Let's build the GPT Tokenizer" Tokenizers are a completely separate stage of the LLM pipeline: they have their own training set training algorithm (Byte Pair Encoding) and after training implement two functions: encode() from strings to tokens and" [X Link](https://x.com/anyuser/status/1760051224043213305) 2024-02-20T21:18Z 15.1K followers, 77.3K engagements "New (2h13m 😅) lecture: "Let's build the GPT Tokenizer" Tokenizers are a completely separate stage of the LLM pipeline: they have their own training set training algorithm (Byte Pair Encoding) and after training implement two functions: encode() from strings to tokens and decode() back from tokens to strings. In this lecture we build from scratch the Tokenizer used in the GPT series from OpenAI" [X Link](https://x.com/anyuser/status/1759996549109776702) 2024-02-20T17:40Z 1.8M followers, 1.7M engagements "Woah. This is the fastest Transformers.js has ever been 🤯 Run a 1.7B LLM 100% locally in your browser at over [---] tokens per second 🚀 No server required. ⚡ WebGPU-accelerated in-browser inference 📦 Optimized ONNX exports w/ GQA 💬 Multi-round conversations w/ KV caching" [X Link](https://x.com/anyuser/status/1914470321450635418) 2025-04-22T00:04Z 15.1K followers, 55.2K engagements "DINOv3 is revolutionary: a new state-of-the-art vision backbone trained to produce rich dense image features. I loved their demo video so much that I decided to re-create their visualization tool. Everything runs 100% in-browser with 🤗 Transformers.js Demo + source code 👇 Introducing DINOv3: a state-of-the-art computer vision model trained with self-supervised learning (SSL) that produces powerful high-resolution image features. For the first time a single frozen vision backbone outperforms specialized solutions on multiple long-standing dense https://t.co/nwS3zFCaaN Introducing DINOv3: a" [X Link](https://x.com/anyuser/status/1956473274583371906) 2025-08-15T21:49Z 15.1K followers, 75.7K engagements "Introducing DINOv3: a state-of-the-art computer vision model trained with self-supervised learning (SSL) that produces powerful high-resolution image features. For the first time a single frozen vision backbone outperforms specialized solutions on multiple long-standing dense prediction tasks. Learn more about DINOv3 here: https://ai.meta.com/blog/dinov3-self-supervised-vision-model/utm_source=twitter&utm_medium=organic_social&utm_content=video&utm_campaign=dinov3" [X Link](https://x.com/anyuser/status/1956027795051831584) 2025-08-14T16:19Z 750.5K followers, 902.1K engagements "Introducing Distil-Whisper Web: 49% smaller 4.2x faster Whisper directly in your browser 🚀 Here is a side-by-side comparison with OpenAI's original version 🤯" [X Link](https://x.com/anyuser/status/1720460890560975103) 2023-11-03T15:20Z 15.1K followers, 273.8K engagements "Depth Anything V2 just released enabling real-time depth estimation directly in your browser with 🤗 Transformers.js and WebGPU acceleration ⚡ The smallest model is only 50MB (@ fp16) making it perfect for on-device usage 😍 Check out the demo (+ source code) 👇" [X Link](https://x.com/anyuser/status/1801672335830798654) 2024-06-14T17:45Z 15.1K followers, 42.6K engagements "WebGPU is the future 😍🔥 Transformers.js can now perform real-time background removal powered by MODNet ⚡ Development for Transformers.js v3 (which adds full WebGPU support) is well underway and we're excited to continue sharing updates and demos Try it out yourself 👇" [X Link](https://x.com/anyuser/status/1768034516134760831) 2024-03-13T22:00Z 15.1K followers, 47.4K engagements "Okay this is insane. WebGPU-accelerated semantic video tracking powered by DINOv3 and Transformers.js 🤯 This will revolutionize AI-powered video editors. which can now run 100% locally in your browser no server inference required (costs $0) 😍 Who's building this" [X Link](https://x.com/anyuser/status/1958893388645753181) 2025-08-22T14:05Z 15.1K followers, 43.7K engagements "IBM just released Granite-4.0 Nano their smallest LLMs ever (300M & 1B) 😍 The models demonstrate remarkable instruction following and tool calling capabilities and can even run locally in-browser This means they can interact with websites and call browser APIs for you 🤯" [X Link](https://x.com/anyuser/status/1983218720366326002) 2025-10-28T17:06Z 15.1K followers, 33.1K engagements "I know it just released but I don't see many people talking about the Phi-3 tokenizer 👀 Here's the full list of added special tokens. what do you notice 🤯 assistant step function_output tag function_call system end raw continue user function_list calc code /code summary resource assistant_mask start message fim_prefix fim_middle fim_suffix meta_start ipynb_marker diff_marker ghissue ghreview disc_start disc_sep disc_threadquery /query data /data sys /sys inst /inst" [X Link](https://x.com/anyuser/status/1782799726443684271) 2024-04-23T15:52Z 15.1K followers, 100.1K engagements "Llama [---] running 100% locally in your browser on WebGPU 🦙 Up to [--] tokens per second ⚡ Powered by 🤗 Transformers.js and ONNX Runtime Web. No installation required. just visit a website Check out the demo and source code below 👇" [X Link](https://x.com/anyuser/status/1840767709317046460) 2024-09-30T14:56Z 15.1K followers, 35.2K engagements "Introducing Kokoro Web: ML-powered speech synthesis directly in your browser 🚀 It's 100% free and supports unlimited generations. Choose from [--] unique voices offering natural and expressive speech for any application. Now w/ streaming & WebGPU acceleration Check it out 👇" [X Link](https://x.com/anyuser/status/1890454815525736567) 2025-02-14T17:35Z 15.1K followers, 37.4K engagements "Introducing Moonshine Web: real-time speech recognition running 100% locally in your browser 🚀 Faster and more accurate than Whisper 🔒 Privacy-focused (no data leaves your device) ⚡ WebGPU accelerated (w/ WASM fallback) 🔥 Powered by ONNX Runtime Web and Transformers.js" [X Link](https://x.com/anyuser/status/1869423057741230539) 2024-12-18T16:42Z 15.1K followers, 28.8K engagements "Massive size reductions for Transformers.js models: - Whisper-tiny decoder: 50MB 30MB (-40%) - NLLB decoder: 732MB 476MB (-35%) - bloom: 819MB 562MB (-31%) - gpt2: 167MB 127MB (-24%) - bert-base: 134MB 111MB (-17%) . all this with ZERO drop in performance 🤯" [X Link](https://x.com/anyuser/status/1698742891118493905) 2023-09-04T17:00Z 15.1K followers, 108K engagements "YOLOv9 just released and now it's compatible with 🤗 Transformers.js That's right. near real-time object detection running locally in your browser: no server required 🤯 Try it out yourself 👇 🔗 Demo: https://hf.co/spaces/Xenova/yolov9-web https://hf.co/spaces/Xenova/yolov9-web" [X Link](https://x.com/anyuser/status/1761096573755302267) 2024-02-23T18:31Z 15.1K followers, 71.1K engagements "When do you *really* need to use a vector database 🤔 To try answer that question I recreated my semantic image search application to run 100% in-browser with Transformers.js (no server). After loading the model and database it only takes 50ms to compute text embeddings and perform similarity search across 25k images. No vector DB just plain old JavaScript. I think @karpathy was onto something. 👀 What do you think 🔗 Demo site: Source code: https://github.com/xenova/transformers.js/tree/main/examples/semantic-image-search-client" [X Link](https://x.com/anyuser/status/1705385934072742015) 2023-09-23T00:57Z 15.1K followers, 186.8K engagements "@sinclanich np.array people keep reaching for much fancier things way too fast these days" [X Link](https://x.com/anyuser/status/1647374645316968449) 2023-04-15T23:01Z 1.8M followers, 663.3K engagements "Introducing SmolLM: a new SOTA series of 135M 360M and 1.7B models perfect for on-device deployment 🔥 We also uploaded ONNX weights for the models meaning they can run locally in your browser with 🤗 Transformers.js and WebGPU acceleration ⚡ Try it out (+ blog post) 👇" [X Link](https://x.com/anyuser/status/1813258097185448377) 2024-07-16T17:03Z 15.1K followers, 34K engagements "Transformers.js v2.0 is finally here 🔥 Run @huggingface transformers directly in your browser with no need for a server Some of the new features include: 🛠 Complete ES6 rewrite 📄 Documentation + examples 🤗 Improved Hugging Face Hub integration https://github.com/xenova/transformers.js https://github.com/xenova/transformers.js" [X Link](https://x.com/anyuser/status/1658893772481019904) 2023-05-17T17:54Z 15.1K followers, 67K engagements "Jina CLIP v1 just released: a new state-of-the-art multimodal embedding model that outperforms OpenAI CLIP in text-image retrieval 😍 We also contributed ONNX weights so it's now compatible with 🤗 Transformers.js v3 and runs with WebGPU acceleration ⚡ Try out the demo 👇" [X Link](https://x.com/anyuser/status/1798321533405847771) 2024-06-05T11:50Z 15.1K followers, 45.3K engagements "Introducing TTS WebGPU: The first ever text-to-speech web app built with WebGPU acceleration 🔥 High-quality and natural speech generation that runs 100% locally in your browser powered by OuteTTS and Transformers.js.🤗 Try it out yourself Demo + source code below 👇" [X Link](https://x.com/anyuser/status/1865841995802882132) 2024-12-08T19:32Z 15.1K followers, 21.3K engagements "WOW 🤯 DINOv3 can run locally on your phone. from the browser This unlocks endless possibilities for AI-powered web apps. 🤏 Model is tiny (only 15MB at 4-bit quantization) 🧠 Delivers powerful high-resolution image features ✨ Works completely offline Try it yourself 👇" [X Link](https://x.com/anyuser/status/1956763976080970071) 2025-08-16T17:04Z 15.1K followers, 40K engagements "I can't believe this. Phi-3.5-mini (3.8B) running in-browser at [--] tokens/second on WebGPU w/ Transformers.js and ONNX Runtime Web 🤯 Since everything runs 100% locally no messages are sent to a server a huge win for privacy 🔒 Check out the demo + source code below 👇" [X Link](https://x.com/anyuser/status/1826992922509595068) 2024-08-23T14:40Z 15.1K followers, 22.3K engagements Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing
@xenovacom Xenova
Xenova posts on X about model, demo, code, check the most. They currently have [---------] followers and [---] posts still getting attention that total [-----] engagements in the last [--] hours.
Engagements: [-----] #
- [--] Week [-------] +5,489%
- [--] Month [-------] +3,423%
- [--] Months [-------] +214%
- [--] Year [-------] -77%
Mentions: [--] #
Followers: [---------] #
- [--] Month [------] +1.80%
- [--] Months [------] +12%
- [--] Year [------] +22%
CreatorRank: [-------] #
Social Influence
Social category influence technology brands stocks social networks events automotive brands finance
Social topic influence model, demo, code, check, ai, image, inference #148, in the, realtime, a new
Top accounts mentioned or mentioned by @huggingface @alibabaqwen @huggingfacetransformers @karpathy @vercels @sinclanich @briaais @simonw @karpathys @next @ldenoue @fleetwood @vikhyatk @alonsosilva @evilpingwin @deepseekai @jinaai @bytedanceoss @clementdelangue @tobi
Top assets mentioned Alphabet Inc Class A (GOOGL) Microsoft Corp. (MSFT) IBM (IBM)
Top Social Posts
Top posts by engagements in the last [--] hours
"@ldenoue Good news I found this issue. It was a single line call to Math.max in the whisper logits processor 🙄 For some reason the "recommended" way to find the maximum element in a list is super error-prone. Pushing a fix now"
X Link 2023-05-29T14:50Z 10.4K followers, [--] engagements
"Transformers.js just hit [----] stars on GitHub 🤯 The #WebML community is growing so fast and I'm proud to be a part of it 🤗 If you ever plan on adding in-browser machine-learning functionality to your website or web-app check out the project:"
X Link 2023-07-08T03:02Z [----] followers, 30.5K engagements
"AI code completion running 100% locally inside your browser thanks to @BigCodeProject's StarCoder models and 🤗 Transformers.js We also got their new 1B model running at [--] tokens per second in Node.js (CPU). 🚀 Check out the demo"
X Link 2023-07-30T16:38Z [----] followers, 21.2K engagements
"Two annoying things about OpenAI's tokenizer playground: (1) it's capped at 50k characters and (2) it doesn't support GPT-4 or GPT-3.5. So I built my own version w/ Transformers.js It can tokenize the entire "Great Gatsby" (269k chars) in 200ms 🚀 https://huggingface.co/spaces/Xenova/the-tokenizer-playground https://huggingface.co/spaces/Xenova/the-tokenizer-playground"
X Link 2023-08-04T19:56Z 15.1K followers, 48.3K engagements
"Since everything runs 100% locally in your browser you can get real-time feedback as you type 🔥 PS: Check out 🤗 Transformers.js if you haven't already:"
X Link 2023-08-04T19:56Z [----] followers, [----] engagements
"@simonpfish I love this - so creative 🔥 We need more emojis though especially 🤗 Also Id be keen to turn this into a web app w/ transformers.js So let me know when you release the code"
X Link 2023-08-05T03:14Z 10.7K followers, [---] engagements
"Here's a sneak peek of my "Chat with YouTube" browser extension made with @Vercel's AI SDK 🔥 It uses Llama-v2 (7B) deployed with @HuggingFace inference endpoints. 🚀 Source code and tutorial coming soon 🤗"
X Link 2023-08-10T16:39Z [----] followers, [----] engagements
"Transformers.js v2.5.2 now supports audio classification w/ MMS and wav2vec2 meaning you can for example perform language identification for over [----] languages 🤯🔥 Get started in just [--] lines of code 👇 Full release notes:"
X Link 2023-08-14T21:44Z [----] followers, 24.1K engagements
"Introducing Chat with YouTube an AI-powered browser extension that lets you chat with YouTube videos 🚀 This project shows how easy it is to build conversational browser extensions using 🤗 Inference Endpoints and @Vercel's AI SDK. + it's open source https://github.com/xenova/chat-with-youtube https://github.com/xenova/chat-with-youtube"
X Link 2023-08-15T22:33Z 15.1K followers, 52.7K engagements
"@smjain All you really need to do is structure your model repo like ours (with all onnx files in an onnx subfolder). You can also use our conversation script to help with this (and it also does quantization):"
X Link 2023-09-05T16:18Z [----] followers, [---] engagements
"WOW 🤯 a GPT token probability visualizer now running client-side with Transformers.js 🤗 Check it out 👇"
X Link 2023-09-05T22:45Z [----] followers, [----] engagements
"We just released Transformers.js v2.6.0 New features: - [--] new architectures: BLOOM MPT BeiT CamemBERT CodeLlama GPT-J mBART ResNet WavLM and more 😍 - Over [---] newly-converted models on the Hub 🚀 - Huge model size reductions (up to -40%) 👇"
X Link 2023-09-08T15:36Z [----] followers, 39.8K engagements
"WebGPU-accelerated automatic speech recognition running completely in your browser 😍 What an amazing job @fleetwood___ 🚀 Go check it out 👇"
X Link 2023-09-12T14:03Z [----] followers, [----] engagements
"WOW 🤯 An in-browser version of ChatGPT (or HF Chat) built with 🤗 Transformers.js Yes that's right everything runs 100% locally in your browser meaning no need for a server Check it out 🔗 https://huggingface.co/spaces/mithril-security/blind_chat 🚀Just released #BlindChat: an open-source & privacy-first ChatGPT alternative 🔒BlindChat combines @huggingface transformers.js (from the great @xenovacom) with #ChatUI for a private and fully in-browser experience Try it on HF: https://t.co/REr4LA0c9r https://huggingface.co/spaces/mithril-security/blind_chat 🚀Just released #BlindChat: an"
X Link 2023-09-21T17:29Z 15.1K followers, 105.6K engagements
"When do you really need to use a vector database 🤔 To try answer that question I recreated my semantic image search application to run 100% in-browser with Transformers.js (no server). After loading the model and database it only takes 50ms to compute text embeddings and perform similarity search across 25k images. No vector DB just plain old JavaScript. I think @karpathy was onto something. 👀 What do you think 🔗 Demo site: Source code: https://github.com/xenova/transformers.js/tree/main/examples/semantic-image-search-client"
X Link 2023-09-23T00:57Z 15.1K followers, 186.8K engagements
"Transformers.js v2.6.2 now supports Document Question Answering meaning you can easily extract information from images. directly in your browser (no server needed) 🤯 We also added new models like Donut LongT5 and Blenderbot 🥳 I can't wait to see what you build 🤗"
X Link 2023-09-27T14:29Z [----] followers, 14.6K engagements
"Which tasks/models should we add next 🤔 Let us know in the comments 👇 #WebML Full release notes:"
X Link 2023-09-27T14:29Z [----] followers, [---] engagements
"@ldenoue Could you open up a GitHub issue Will make debugging and helping a lot easier 🤗"
X Link 2023-10-03T19:40Z [----] followers, [--] engagements
"Woah 🤯 A new 20M parameter embeddings model that gives similar performance to OpenAI's text-embedding-ada-002 but is much smaller+faster 🔥 I don't understand why 95% of developers still use closed-source embeddings models. 😅 + it's compatible with 🤗 Transformers.js New embeddings model gte-tiny is published Distilled from gte-small offering slightly-worse performance with half the layers. (Alternatively same size but better performance compared to all-MiniLM-L6-v2.) ONNX models also available. Check it out (link below) https://t.co/ogARt355Ne New embeddings model gte-tiny is published"
X Link 2023-10-06T17:34Z 15.1K followers, 102.8K engagements
"We just released 🤗 Transformers.js v2.7.0 which adds supports for 🗣 Text to Speech w/ speecht5. This means you can now synthesize human speech directly in your browser. no server required 🔥 Check out the demo 👇 https://huggingface.co/spaces/Xenova/text-to-speech-client https://huggingface.co/spaces/Xenova/text-to-speech-client"
X Link 2023-10-23T16:07Z 15.1K followers, 66.4K engagements
"A new open-source embeddings model with 8K context length that matches the performance of text-embedding-ada-002 🤯 This is a game changer 🔥 And now it's compatible with 🤗 Transformers.js meaning you can generate embeddings in your browser Node.js or even Deno Introducing jina-embeddings-v2 the world's first open-source model boasting an 8K context length. Matching the prowess of OpenAI's proprietary models now accessible on @huggingface signaling a significant milestone in the landscape of text embeddings. https://t.co/E4fRgTLsR4 Introducing jina-embeddings-v2 the world's first open-source"
X Link 2023-10-27T14:02Z 15.1K followers, 36.2K engagements
"Introducing Distil-Whisper Web: 49% smaller 4.2x faster Whisper directly in your browser 🚀 Here is a side-by-side comparison with OpenAI's original version 🤯"
X Link 2023-11-03T15:20Z 15.1K followers, 273.8K engagements
"Yes that's right. the new Distil-Whisper models from @huggingface are fully compatible with Transformers.js 🤗 This means you can generate high-quality transcripts directly in JavaScript: in-browser Node or even Deno 🤯🚀 Get started in just [--] lines of code:"
X Link 2023-11-04T21:32Z [----] followers, 41.4K engagements
"This is an absolute game changer 🤯 @threejs but for Gaussian Splatting 🔥 I cant wait to see what the web-dev community builds with this 🤗 cc @mrdoob"
X Link 2023-11-06T17:10Z [----] followers, 26.2K engagements
"We just released 🤗 Transformers.js v2.8.0 which adds a ton of new features including: 🖼 Super-resolution and image restoration w/ Swin2SR ✍ Optical character recognition w/ TrOCR 💬 Text-generation w/ Mistral and Falcon (1B params) More details in 🧵👇"
X Link 2023-11-09T17:04Z 15.1K followers, 46.6K engagements
"Transformers.js just hit [--] million total requests on @jsDelivr with 52% of them coming in the past [--] days alone 🤯 We have a ton of exciting updates coming soon so stay tuned I'm excited to show you what's next. 🚀🔥"
X Link 2023-11-12T18:13Z [----] followers, 39.3K engagements
"🤗 Transformers.js just hit 5K stars on GitHub 🌟 Thank you to everyone in the community for your support and contributions. this is why open source is the best 🔥 PS: Stay tuned for some exciting updates coming soon 🚀"
X Link 2023-12-02T13:51Z [----] followers, 10.7K engagements
"Say goodbye to silent performance issues when prompting LLMs Today we released 🤗 Transformers.js v2.12 which adds support for chat templating 💬 This means you can generate LLM inputs for almost any model on the @huggingface Hub directly in your browser w/ JavaScript 🤯"
X Link 2023-12-19T00:28Z [----] followers, [----] engagements
"🚨 We're kicking off [----] with several improvements for Transformers.js developers: - Conditional typing of pipelines based on task. - Inline documentation + code snippets. - Pipeline-specific call parameters and return types. What should we add next Let us know 🤗👇"
X Link 2024-01-03T15:00Z [--] followers, [----] engagements
"Image-to-LaTeX in [--] lines of JavaScript code with 🤗 Transformers.js This is made possible thanks to @vikparuchuri's amazing texify2 model which we converted to ONNX so it can run in the browser 🔥 💡 Project idea: browser extension to convert PDFs/screenshots to LaTeX"
X Link 2024-01-14T11:47Z [----] followers, 12.6K engagements
"@VikParuchuri Yes I noticed that so I made a custom ONNX export config with Optimum 🔥 Validation tests passed with 1e-5 atol on the logits (randomised inputs)"
X Link 2024-01-14T16:10Z [----] followers, [--] engagements
"Introducing the 🛝 Jinja Playground: Design LLM chat templates directly in your browser with instant feedback. Built with @huggingface/jinja a minimalistic JavaScript implementation of the Jinja templating engine specifically designed for parsing + rendering chat templates"
X Link 2024-01-23T11:29Z [----] followers, 18.5K engagements
"🚨 Hugging Chat Assistants are out 🚨 Just like OpenAI's GPTs you can now create your own personal assistant in Hugging Chat 🤯 To test it out I built a "Prisoner Interrogation" game where you must try to extract a secret password from a prisoner. Can you do it 🤔"
X Link 2024-02-02T15:13Z [----] followers, 26.2K engagements
"Local background removal Figma plugin built with 🤗 Transformers.js and BRIA AI's RMBG-v1.4 model This shows what an amazing opportunity it is for JavaScript developers to build powerful AI applications without worrying about API/server costs Great work @enzostvs 🔥"
X Link 2024-02-19T14:25Z [----] followers, 14.7K engagements
"@RemiCadene @Tesla Welcome to the team 🤗🤗🤗"
X Link 2024-03-07T15:12Z [----] followers, [---] engagements
"We just updated our in-browser Background Removal demo to use WebGPU and it's now 50x faster 🤯 [--] seconds down to 180ms ⚡ Powered by @bria_ai_'s RMBG-v1.4 model and 🤗 Transformers.js . and yes the video is in real time 🤯"
X Link 2024-03-09T21:16Z 15.1K followers, 76.6K engagements
"@anotherjesse This is so cool 🔥 You should try out the WebGPU version (I've tested and it should be compatible with CLIP). You can try it out here: (see first comment for installation and running instructions)"
X Link 2024-03-16T21:06Z [----] followers, [---] engagements
"Grok-1 is finally out 😍 But while everyone was focused on the weights I decided to take a look at the tokenizer. I also added it to the Tokenizer Playground Structurally it looks quite similar to the Llama [--] tokenizer (BPE w/ byte-fallback) with a vocabulary size of [--] = [------] (much larger than Llama's 32k). add_prefix_space=True so "hello world" becomes [-----] [----] // "hello" "world" This is the same as Llama but different to GPT and Gemma. In addition to the "PAD" "BOS" "EOS" and "UNK" special tokens there are "separator" and "mask" tokens as well as [--] control tokens of the form "controlX""
X Link 2024-03-18T02:07Z 15.1K followers, 30.5K engagements
"New features in 🤗 Transformers.js v2.16.1: 🔥 New models: APISR for Anime Super-Resolution and EfficientNet for image classification 🖼 New pipeline: Image Feature Extraction 💬 Improved chat templating support: C4AI Command-R tool and RAG prompt generation See 🧵 for more info"
X Link 2024-03-20T17:46Z 15.1K followers, 26.1K engagements
"Snowflake just released Arctic Embed a collection of open-source text embedding models optimized for retrieval accuracy and efficiency ❄ 📄 Apache [---] license 🌐 Great for in-browser use w/ 🤗 Transformers.js (22 [---] M params) ⚡ WebGPU-accelerated (120x faster than WASM)"
X Link 2024-04-18T10:05Z [----] followers, [----] engagements
"Meta's Llama [--] is here with a brand new tokenizer 🦙 I've added it to the Tokenizer Playground so you can experiment with it in your browser: For those interested here are the key differences over Llama 2: [--]. 4x larger vocabulary (32K - 128K). This means improved token efficiency (up to 15% fewer tokens compared to Llama 2) as well as stronger multilingualism. However it comes at the cost of much larger input and output embedding matrices which account for a sizeable portion of the parameter count increase of the small model going from 7B in Llama [--] to 8B in Llama [--]. [--]. Llama [--] uses a"
X Link 2024-04-18T16:37Z [----] followers, 12.8K engagements
"Meta's Segment Anything Model (SAM) can now run in your browser w/ WebGPU (+ fp16) meaning up to 8x faster image encoding (10s 1.25s) 🤯⚡ Video is not sped up Everything runs 100% locally thanks to 🤗 Transformers.js and onnxruntime-web 🔗 Demo: https://hf.co/spaces/Xenova/segment-anything-webgpu https://hf.co/spaces/Xenova/segment-anything-webgpu"
X Link 2024-04-21T00:44Z 15.1K followers, 122.4K engagements
"Introducing Phi-3 WebGPU a private and powerful AI chatbot that runs locally in your browser powered by 🤗 Transformers.js and onnxruntime-web 🔒 On-device inference: no data sent to a server ⚡ WebGPU-accelerated ( [--] t/s) 📥 Model downloaded once and cached Try it out 👇"
X Link 2024-05-08T12:00Z 15.1K followers, 119K engagements
"Phi-3 running at [--] tokens per second 100% locally in your browser 🤯⚡ What speed do you get Try it out yourself 👇 https://huggingface.co/spaces/Xenova/experimental-phi3-webgpu https://huggingface.co/spaces/Xenova/experimental-phi3-webgpu The always brilliant @xenovacom released an experimental WebGPU powered demo of phi-3 running directly in the browser. My consumer but well-specced workstation got [--] tokens per second The age of usable on-device inference is rapidly approaching https://t.co/1ve1FGAoBd https://huggingface.co/spaces/Xenova/experimental-phi3-webgpu"
X Link 2024-05-09T20:15Z [----] followers, 19.2K engagements
"@magikarp_tokens This is because Llama-3 switched from sentencepiece BPE to tiktoken BPE which actually introduced many illegal tokens (those that cant be reached using the merges according to the standard BPE algorithm). v0.19.1 fixed this here: More info here: https://github.com/huggingface/tokenizers/pull/1493 So what's the deal with tiktoken vs. sentencepiece Surely the BPE algorithm is universal Unfortunately it's not that simple. (image 1) The difference can be traced back to this section of code in tiktoken: https://t.co/ujKC3LZg94 At first it looks like an optimization to"
X Link 2024-05-12T11:00Z [----] followers, [---] engagements
"The GPT-4o tokenizer is so much better than GPT-4 for non-English languages 🤯 Here's a demo across [--] different languages: [----] [---] (-52%) tokens for the exact same text (1660 characters) 🔥 GPT-4o (left) vs. GPT-4 (right)"
X Link 2024-05-13T21:02Z 15.1K followers, 36.9K engagements
"WOW I can't believe that Transformers.js was mentioned on stage at Google I/O 🤯 This is a dream come true 🤗"
X Link 2024-05-14T22:38Z 15.1K followers, 29.2K engagements
"Moondream your favorite tiny vision language model by @vikhyatk can now run directly in the browser on WebGPU 🤯 Powered of course by Transformers.js and ONNX Runtime Web 🤗 Local inference means no data leaves your device This is huge for privacy Try it out yourself 👇"
X Link 2024-05-17T11:53Z 15.1K followers, 38.5K engagements
"You can now use 🤗 Transformers.js with Google Visual Blocks a visual programming framework that lets you create machine learning pipelines in a no-code graph editor 🛠 Rapid workflow prototyping 💡 Intuitive drag-and-drop interface ⚡ WebGPU-accelerated in-browser inference"
X Link 2024-05-20T14:59Z 15.1K followers, 22.7K engagements
"Mistral-7B-Instruct-v0.3 is out and now supports function calling It has an updated tokenizer so we added it to The Tokenizer Playground. 🔥 If you're learning about tokenization needing to count tokens or just want to play around with it then check out our web demo 🤗"
X Link 2024-05-22T21:04Z [----] followers, [---] engagements
"Transformers.js is being added to Firefox [---] 🤯 Thats right fully private on-device AI directly in your browser 🔥 The first use-case theyre exploring is automatic alt-text generation for images. A huge win for accessibility What other features would you like to see 👇 Firefox [---] is bringing a game-changing feature: automatic alt-text generation for images using a fully private on-device AI model 🙌🏾 Initially available in the built-in PDF editor our aim is to extend this to general browsing for screen reader users. https://t.co/019T4CbcW6 Firefox [---] is bringing a game-changing feature:"
X Link 2024-06-02T15:14Z 15.1K followers, 44.7K engagements
"@jpohhhh @realmrfakename @cartesia_ai Should be possible. 👀 Others have been able to create working ONNX versions of popular diffusion models like stable diffusion so I don't see what's stopping you from getting StyleTTS working :) Keep me updated on your progress"
X Link 2024-06-04T22:51Z [----] followers, [--] engagements
"Jina CLIP v1 just released: a new state-of-the-art multimodal embedding model that outperforms OpenAI CLIP in text-image retrieval 😍 We also contributed ONNX weights so it's now compatible with 🤗 Transformers.js v3 and runs with WebGPU acceleration ⚡ Try out the demo 👇"
X Link 2024-06-05T11:50Z 15.1K followers, 45.3K engagements
"It's finally possible: real-time in-browser speech recognition with OpenAI Whisper 🤯 The model runs fully on-device using Transformers.js and ONNX Runtime Web and supports multilingual transcription across [---] different languages 🔥 Check out the demo (+ source code) 👇"
X Link 2024-06-07T16:05Z 15.1K followers, 266.9K engagements
"@tombielecki Definitely possible - it would just require updating the initial tokens passed to the decoder. Do you have an example (in python) for this I can take a look at Feel free to open a feature request on GitHub so I can track this easier. https://github.com/xenova/transformers.js https://github.com/xenova/transformers.js"
X Link 2024-06-07T23:57Z [----] followers, [----] engagements
"@whitphx @maartenbreddels @alonsosilva @Alibaba_Qwen @solara_dev It certainly can 🚀 Source code: Demo: https://github.com/xenova/transformers.js/tree/v3/examples/webgpu-chat Qwen2 is here with base and instruct models released across [--] sizes: 0.5B 1.5B 7B 57B-A14B and 72B 😍 To test them out I created a WebGPU chat demo of the 0.5B version with 🤗 Transformers.js It runs 100% locally in the browser (no server required). Try it out 👇 https://t.co/PIOQ2CTzPw https://github.com/xenova/transformers.js/tree/v3/examples/webgpu-chat Qwen2 is here with base and instruct models released across 5"
X Link 2024-06-10T12:02Z [----] followers, [---] engagements
"Depth Anything V2 just released enabling real-time depth estimation directly in your browser with 🤗 Transformers.js and WebGPU acceleration ⚡ The smallest model is only 50MB (@ fp16) making it perfect for on-device usage 😍 Check out the demo (+ source code) 👇"
X Link 2024-06-14T17:45Z 15.1K followers, 42.6K engagements
"Florence-2 the new vision foundation model by Microsoft can now run 100% locally in your browser on WebGPU thanks to Transformers.js 🤗🤯 It supports tasks like image captioning optical character recognition object detection and many more 😍 WOW Demo (+ source code) 👇"
X Link 2024-06-26T15:42Z 15.1K followers, 90.6K engagements
"After my recent post about running Gemini Nano locally in your browser using Chrome's new window.ai feature many people asked how to set it up themselves. 🤔 So I wrote a blog post/tutorial explaining the process 🔥 Hope it helps 👇 https://huggingface.co/blog/Xenova/run-gemini-nano-in-your-browser https://huggingface.co/blog/Xenova/run-gemini-nano-in-your-browser"
X Link 2024-07-11T15:53Z [----] followers, [----] engagements
"Introducing SmolLM: a new SOTA series of 135M 360M and 1.7B models perfect for on-device deployment 🔥 We also uploaded ONNX weights for the models meaning they can run locally in your browser with 🤗 Transformers.js and WebGPU acceleration ⚡ Try it out (+ blog post) 👇"
X Link 2024-07-16T17:03Z 15.1K followers, 34K engagements
"Mistral and NVIDIA just released Mistral NeMo a state-of-the-art 12B model with 128k context length 😍 It uses a new Tiktoken-based tokenizer which is far more efficient at compressing source code and non-English languages (3x for Korean 🤯) Tokenizer playground link 👇"
X Link 2024-07-18T16:07Z [----] followers, [----] engagements
"Introducing Whisper Diarization: Multilingual speech recognition with word-level timestamps and speaker segmentation running 100% locally in your browser thanks to 🤗 Transformers.js Tested on this iconic Lettermen interview w/ Grace Hopper from [----] Demo (+ source code) 👇"
X Link 2024-07-22T16:47Z 15.1K followers, 94.9K engagements
"🤏 SmolLM Instruct v0.2 just dropped: a series of blazingly fast and remarkably powerful small language models (135M 360M and 1.7B params). Apache [---] licensed 🔥 They can even run in-browser on WebGPU with 🤗 Transformers.js at [--] tokens/second 🤯 Demo (+ source code) 👇"
X Link 2024-08-20T14:09Z 15.1K followers, 16.9K engagements
"I can't believe this. Phi-3.5-mini (3.8B) running in-browser at [--] tokens/second on WebGPU w/ Transformers.js and ONNX Runtime Web 🤯 Since everything runs 100% locally no messages are sent to a server a huge win for privacy 🔒 Check out the demo + source code below 👇"
X Link 2024-08-23T14:40Z 15.1K followers, 22.3K engagements
"@nmstoker Running on an RTX [----] courtesy of @evilpingwin 🤗 (who can provide more details if youd like)"
X Link 2024-08-23T14:54Z 10.6K followers, [---] engagements
"There has been a huge debate recently about the best approach for image background removal. Here's my attempt: - In-browser inference w/ 🤗 Transformers.js - WebGPU accelerated (fast) - Costs $0 (no image hosting or server processing) - No data leaves your device (privacy) It works. It's fast. It's super high quality. It's cheap. And it scales. Available for purchase soon https://t.co/5If4T3Inj3 It works. It's fast. It's super high quality. It's cheap. And it scales. Available for purchase soon https://t.co/5If4T3Inj3"
X Link 2024-08-26T17:07Z 15.1K followers, 422.3K engagements
"BONUS: We're releasing the entire project under the Apache [---] license so feel free to adapt the demo and use it for commercial purposes The background removal model (MODNet) is also Apache [---] 🤗 Code: 🔗 Demo: https://huggingface.co/spaces/webml-community/remove-background-webgput=0 https://github.com/huggingface/transformers.js-examples/tree/main/remove-background-webgpu https://huggingface.co/spaces/webml-community/remove-background-webgput=0 https://github.com/huggingface/transformers.js-examples/tree/main/remove-background-webgpu"
X Link 2024-08-27T21:56Z 10.6K followers, [----] engagements
"@quicksave2k @fleetwood___ Indeed well work with the ONNX Runtime Web team to get this added 🤗🔥 @fleetwood___ Do you have plans to add support for them in Transformers.js I assume it means adding them to ONNX runtime web first right @fleetwood___ Do you have plans to add support for them in Transformers.js I assume it means adding them to ONNX runtime web first right"
X Link 2024-09-24T02:35Z 10.6K followers, [--] engagements
"Llama [---] running 100% locally in your browser on WebGPU 🦙 Up to [--] tokens per second ⚡ Powered by 🤗 Transformers.js and ONNX Runtime Web. No installation required. just visit a website Check out the demo and source code below 👇"
X Link 2024-09-30T14:56Z 15.1K followers, 35.2K engagements
"Run OpenAI's new Whisper Turbo model 100% locally in your browser with Transformers.js ⚡ Transcribe [--] minutes of audio in [--] seconds 🤯 Demo + source code 👇"
X Link 2024-10-01T11:37Z 15.1K followers, 141.7K engagements
"Google just released Gemma-2-JPN a fine-tuned version of Gemma [--] 2B on Japanese text. It supports the Japanese language at the same performance as English-only queries on Gemma [--] 😍 You can even run it 100% locally in your browser on WebGPU using 🤗 Transformers.js 👇 GoogleGemma-2-JPNGemma [--] 2Bfine-tuneGemma [--] : https://t.co/LgyHREpCOf GoogleGemma-2-JPNGemma [--] 2Bfine-tuneGemma [--] : https://t.co/LgyHREpCOf"
X Link 2024-10-03T11:06Z 10.6K followers, [----] engagements
"Google is on fire with their open source releases 🔥 Today they dropped Gemma-APS a collection of Gemma models for text-to-propositions segmentation. The models are distilled from fine-tuned Gemini Pro models applied to multi-domain synthetic data 👇 https://huggingface.co/collections/google/gemma-aps-release-66e1a42c7b9c3bd67a0ade88 https://huggingface.co/collections/google/gemma-aps-release-66e1a42c7b9c3bd67a0ade88"
X Link 2024-10-15T16:03Z 15.1K followers, 41.1K engagements
"After more than a year of development we're excited to announce the release of 🤗 Transformers.js v3 ⚡ WebGPU support (up to 100x faster than WASM) 🔢 New quantization formats (dtypes) 🏛 [---] supported architectures in total 📂 [--] new example projects and templates 🤖 Over [----] pre-converted models 🌐 Node.js (ESM + CJS) Deno and Bun compatibility 🏡 A new home on GitHub and NPM Get started with npm i @huggingface/transformers. Learn more in the blog post below 👇"
X Link 2024-10-22T15:02Z 15.1K followers, 44.7K engagements
"@NERDDISCO @huggingface @Microsoft @v0 Awesome 🤩🙌"
X Link 2024-10-29T09:21Z 13.9K followers, [---] engagements
"OmniParser the new screen parsing tool from Microsoft (and #1 trending model on @huggingface) can now run 100% locally in your browser with Transformers.js 🤯 Who's going to be the first to turn this into a browser extension 👀 Endless possibilities Demo & code below 👇"
X Link 2024-11-01T07:29Z 15.1K followers, 66.2K engagements
"Given a screenshot of a user interface it detects interactable icons/elements on the page (w/ bounding boxes) and generates meaningful descriptions for each. It's meant to enhance the ability of a multimodal model (like GPT-4V) to generate actions (so it can interact with the page). Links to learn more: - GitHub repo: - Blog post: - Project page: https://microsoft.github.io/OmniParser/ https://www.microsoft.com/en-us/research/articles/omniparser-for-pure-vision-based-gui-agent/ https://github.com/microsoft/OmniParser https://microsoft.github.io/OmniParser/"
X Link 2024-11-01T07:51Z 10.7K followers, [----] engagements
"WOW 🤯 Language models are becoming smaller and more capable than ever Here's SmolLM2 running 100% locally in-browser w/ WebGPU on a 6-year-old GPU. Look at that speed ⚡😍 Powered by 🤗 Transformers.js and ONNX Runtime Web How many tokens/second do you get Let me know 👇"
X Link 2024-11-27T13:04Z 15.1K followers, 12.9K engagements
"We just released Transformers.js v3.1 and you're not going to believe what's now possible in the browser w/ WebGPU 🤯 Let's take a look: 🔀 Janus from @deepseek_ai for unified multimodal understanding and generation (Text-to-Image and Image-Text-to-Text) 👁 Qwen2-VL from @alibaba_qwen for dynamic-resolution image understanding 🔢 JinaCLIP from @JinaAI_ for general-purpose multilingual multimodal embeddings 🌋 LLaVA-OneVision from @ByteDanceOSS for Image-Text-to-Text generation 🤸♀ ViTPose for pose estimation 📄 MGP-STR for optical character recognition (OCR) 📈 PatchTST & PatchTSMixer for"
X Link 2024-11-28T14:37Z 15.1K followers, 20.7K engagements
"⚡ Janus WebGPU demo (+ source code): 📝 Transformers.js v3.1 release notes: https://github.com/huggingface/transformers.js/releases/tag/3.1.0 https://huggingface.co/spaces/webml-community/Janus-1.3B-WebGPU https://github.com/huggingface/transformers.js/releases/tag/3.1.0 https://huggingface.co/spaces/webml-community/Janus-1.3B-WebGPU"
X Link 2024-11-28T14:38Z 10.8K followers, [----] engagements
"@NielsRogge @tumble_wood @GozukaraFurkan The WebLLM team have been able to run Llama 70B on WebGPU so it should technically be possible 👀"
X Link 2024-12-03T11:26Z 11.1K followers, [---] engagements
"@javierluraschi Whoops I completely misread your first message. I think I assumed TTS since the original post was about TTS. For STT whisper is definitely my go-to. It comes in different sizes (ranging from 40M params to 1.5B). https://x.com/xenovacom/status/1841080068593905960 Run OpenAI's new Whisper Turbo model 100% locally in your browser with Transformers.js ⚡ Transcribe [--] minutes of audio in [--] seconds 🤯 Demo + source code 👇 https://t.co/Xj0sxwWeCX https://x.com/xenovacom/status/1841080068593905960 Run OpenAI's new Whisper Turbo model 100% locally in your browser with Transformers.js"
X Link 2024-12-09T15:22Z 10.9K followers, [--] engagements
"🚨 We're hiring an intern to join the WebML team @huggingface 🚨 If you're passionate about open source and want to help build JavaScript libraries that empower the next generation of web developers we'd love to hear from you 🤗 Links to apply in 🧵"
X Link 2024-12-09T17:03Z 11K followers, 14.1K engagements
"Europe Middle East and Africa (EMEA) Remote https://apply.workable.com/huggingface/j/656D794A0C/ https://apply.workable.com/huggingface/j/656D794A0C/"
X Link 2024-12-09T17:03Z 10.9K followers, [---] engagements
"Introducing Moonshine Web: real-time speech recognition running 100% locally in your browser 🚀 Faster and more accurate than Whisper 🔒 Privacy-focused (no data leaves your device) ⚡ WebGPU accelerated (w/ WASM fallback) 🔥 Powered by ONNX Runtime Web and Transformers.js"
X Link 2024-12-18T16:42Z 15.1K followers, 28.8K engagements
"After more than [--] years we finally have a replacement for BERT 🤯 Introducing ModernBERT a family of state-of-the-art encoder-only models with 8K sequence length better downstream performance (classification retrieval) and faster processing. An absolute game-changer 👇"
X Link 2024-12-19T16:50Z 11K followers, [----] engagements
"@thenameless7741 @livekit Wow this looks great The model appears to be llama-based so indeed it should be compatible with Transformers.js (after some minor repo restructuring). Let me make a PR https://huggingface.co/livekit/turn-detector https://huggingface.co/livekit/turn-detector"
X Link 2024-12-21T21:09Z 11K followers, [--] engagements
"Is this the future of AI browser agents 👀 WebGPU-accelerated reasoning LLMs are now supported in Transformers.js 🤯 Here's MiniThinky-v2 (1B) running 100% locally in the browser at [--] tps (no API calls) I can't wait to see what you build with it Demo + source code in 🧵👇"
X Link 2025-01-10T12:08Z 15.1K followers, 83.5K engagements
"For the AI builders out there: imagine what could be achieved with a browser extension that (1) uses a powerful reasoning LLM (2) runs 100% locally & privately and (3) can directly access/manipulate the DOM 👀 🧑💻 Code: 🔗 Demo: https://huggingface.co/spaces/webml-community/llama-3.2-reasoning-webgpu https://github.com/huggingface/transformers.js-examples/tree/main/llama-3.2-reasoning-webgpu https://huggingface.co/spaces/webml-community/llama-3.2-reasoning-webgpu https://github.com/huggingface/transformers.js-examples/tree/main/llama-3.2-reasoning-webgpu"
X Link 2025-01-10T12:08Z 11.3K followers, [----] engagements
"@ClementDelangue @tobi Literally working on this right now 😂"
X Link 2025-01-13T17:35Z 11.4K followers, [----] engagements
"@ClementDelangue @tobi I've already uploaded Transformers.js-compatible checkpoints to the HF hub: and now I'm working on a JS implementation of the phonemizer needed to get it running in the browser: https://github.com/xenova/phonemizer.js https://huggingface.co/onnx-community/Kokoro-82M-ONNX https://github.com/xenova/phonemizer.js https://huggingface.co/onnx-community/Kokoro-82M-ONNX"
X Link 2025-01-13T17:37Z 11.4K followers, [----] engagements
"Introducing Kokoro.js a new JavaScript library for running Kokoro TTS an [--] million parameter text-to-speech model 100% locally in the browser w/ WASM. Powered by 🤗 Transformers.js. WebGPU support coming soon 👉 npm i kokoro-js 👈 Link to demo (+ sample code) in 🧵"
X Link 2025-01-16T14:52Z 15.1K followers, 42.6K engagements
"Introducing SmolVLM 256M (& 500M): The worlds smallest multimodal model. Designed for efficiency and perfect for on-device applications 🔥 Its so small it can even run 100% locally in your browser on WebGPU 🤏 Powered by Transformers.js ⚡ Try it out yourself 👇"
X Link 2025-01-23T14:31Z 15.1K followers, 25.9K engagements
"Reasoning models like o3 and o4-mini are advancing faster than ever but imagine what will be possible when they can run locally in your browser 🤯 Well with 🤗 Transformers.js you can do just that Here's Zyphra's new ZR1 model running at over [---] tokens/second on WebGPU ⚡"
X Link 2025-04-16T18:50Z 12.7K followers, [----] engagements
"I seriously cannot believe this is a 0.6B LLM 🤯 @Alibaba_Qwen just released Qwen3 a series of hybrid reasoning models that allow you to control how much "thinking" the model does for a given task. They can even run locally in your browser on WebGPU with 🤗 Transformers.js"
X Link 2025-04-29T01:46Z 15.1K followers, 76.4K engagements
"Let's go 😍 @Alibaba_Qwen just released Qwen3-Embedding a new series of embedding models: 🏆 SOTA performance on MMTEB MTEB and MTEB-Code 📏 Three different sizes (0.6B / 4B / 8B) 🌍 Multilingual (119 languages) 💻 Can run in-browser w/ Transformers.js (+ WebGPU acceleration)"
X Link 2025-06-06T20:14Z 15.1K followers, 50.9K engagements
"Browsers are the perfect place for hybrid AI inference combining the power of cloud AI with the versatility and privacy of local models. ⚡ Dia already has great WebGPU support so I'm looking forward to seeing more on-device models being integrated directly into the browser itself. especially smaller task-specific models for TTS and background removal. Great work @joshm and team I'm excited to see what's next for @diabrowser 🔥 We are floored by the reaction to @diabrowser. Thank you everyone 🥹 Theres so much more to do is an understatement. But I wanted to pause to extend heartfelt gratitude"
X Link 2025-06-17T21:51Z 13.2K followers, [----] engagements
"You can now run our new SmolLM3-3B model directly in your browser on WebGPU ⚡ 🏆 SoTA for its size 🧠 Dual mode reasoning 🌍 Multilingual (English French Spanish German Italian Portuguese) 🤗 Fully open source (model data code & recipes) Try it out yourself 👇 We just released the best 3B model 100% open-source open dataset architecture details exact data mixtures and full training recipe including pre-training mid-training post-training and synthetic data generation for everyone to train their own. Let's go open-source AI We just released the best 3B model 100% open-source open dataset"
X Link 2025-07-11T22:37Z 13.2K followers, 10.2K engagements
"@nic_o_martin @huggingface Exciting times ahead 🚀😍"
X Link 2025-07-21T15:55Z 13.2K followers, [---] engagements
"Did you know you can vibe code Transformers.js web-apps with @_akhaliq's Anycoder 🤯 Here's an AI image describer powered by Moondream running locally in-browser with WebGPU acceleration and deployed to Hugging Face spaces in one click ⚡ Try it out yourself 👇 Kimi K2 + @GroqInc vibe coding in anycoder build a transformers.js app for Xenova/moondream2 model card info and example included in prompt app was deployed successfully on HF in one click https://t.co/hNW2Diuihx Kimi K2 + @GroqInc vibe coding in anycoder build a transformers.js app for Xenova/moondream2 model card info and example"
X Link 2025-07-21T17:18Z 13.3K followers, 15.5K engagements
"A community member trained a tiny Llama model (23M parameters) on [--] million high-quality @lichess games then deployed it to run entirely in-browser with 🤗 Transformers.js Super cool 🔥 It has an estimated ELO of [----]. can you beat it 👀 (runs on both mobile and desktop)"
X Link 2025-07-22T16:24Z 13.9K followers, [----] engagements
"@alonsosilva @lichess Pretty much yes The highest-scoring legal move is selected on each turn. (in this case the creator just performs a single forward pass for every turn so a logits processor isn't used directly. but it's the same principle)"
X Link 2025-07-22T23:19Z 13.9K followers, [---] engagements
"Transformers.js just crossed 250K weekly downloads on NPM 🚀🤯 Huge thanks to the community for your continued support it's been incredible to see what you've built so far 🤗 Big updates coming soon. 👀"
X Link 2025-07-28T23:44Z 13.3K followers, [----] engagements
"We've made significant optimizations to our ONNX export process enabling 2x faster inference for BERT models on WebGPU ⚡ 📉 Before: decomposed operations (slow) 📈 After: unified Multi-Head Attention node (fast) If you're using Transformers.js or ONNX Runtime try it out 🔥"
X Link 2025-08-01T20:13Z 13.4K followers, [----] engagements
"HUGE: OpenAI just released GPT OSS on Hugging Face 🤯 Here's what you need to know: [--]. Two models: gpt-oss-20b and gpt-oss-120b designed for powerful reasoning agentic tasks and more [--]. Mixture-of-experts (MoE) architecture: 21B and 117B total parameters with 3.6B and 5.1B active parameters respectively. [--]. 4-bit quantization scheme (MXFP4) only applied on the MoE weights allowing the 20GB to fit in a single 16GB GPU and the 120B to fit in a single 80GB GPU. [--]. Text-only reasoning models with chain-of-thought and adjustable reasoning effort levels. [--]. Instruction-tuned with support for tool"
X Link 2025-08-05T17:10Z 13.9K followers, [----] engagements
"Did you know you can try the new gpt-oss models (120b & 20b) directly from their model cards 🤯 Look at that speed 🚀 Check it out 👇"
X Link 2025-08-05T18:21Z 13.3K followers, [----] engagements
"gpt-oss-20b: gpt-oss-120b: https://huggingface.co/openai/gpt-oss-120b https://huggingface.co/openai/gpt-oss-20b https://huggingface.co/openai/gpt-oss-120b https://huggingface.co/openai/gpt-oss-20b"
X Link 2025-08-05T18:22Z 13.3K followers, [---] engagements
"gpt-oss-120b (reasoning: high) absolutely crushes the "bouncing ball in a spinning hexagon" test. Try it out in the gpt-oss playground: http://gpt-oss.com http://gpt-oss.com"
X Link 2025-08-05T18:55Z 13.4K followers, 23.8K engagements
"@HCSolakoglu Official demo from OpenAI powered by Hugging Face's Inference Providers 🤗 https://huggingface.co/docs/inference-providers https://huggingface.co/docs/inference-providers"
X Link 2025-08-05T19:09Z 13.3K followers, [---] engagements
"@simonw @lmstudio @ollama @OpenRouterAI @CerebrasSystems Was trying this as you put out your tweet Prompt: "Generate an SVG of a pelican riding a bicycle" Adding "detailed" to the prompt produces a bicycle with spokes. pretty cool"
X Link 2025-08-05T20:39Z 13.3K followers, [---] engagements
"@simonw @lmstudio @ollama @OpenRouterAI @CerebrasSystems (this was the 120b running with reasoning: high)"
X Link 2025-08-05T20:41Z 13.3K followers, [---] engagements
"There's a new tiny TTS model in town: Kitten TTS 🐱 With just 15M parameters (25 MB) it delivers impressive quality for its size and can even run in real time without a GPU. So I created a web demo for it: featuring text normalization chunking and real-time playback. 🤗 Introducing Kitten TTS a SOTA tiny text-to-speech model - Just 15M parameters - Runs without a GPU - Model size less than [--] MB - Multiple high-quality voices - Ultra-fast - even runs on low-end edge devices Github and HF links below https://t.co/9T3u1M0WGo Introducing Kitten TTS a SOTA tiny text-to-speech model - Just 15M"
X Link 2025-08-06T22:50Z 15.1K followers, 14.1K engagements
"Google just released their smallest Gemma model ever: Gemma [--] 270M 🤯 🤏 Highly compact & efficient 🤖 Strong instruction-following capabilities 🔧 Perfect candidate for fine-tuning It's so tiny that it can even run 100% locally in your browser with Transformers.js 🤗"
X Link 2025-08-14T16:15Z 15.1K followers, 35.4K engagements
"@simonw Haha thanks 😁 We dont currently support in-browser fine-tuning but youre not the first (and certainly not the last) to ask about it 👀 Maybe one day 😅"
X Link 2025-08-14T17:36Z 13.4K followers, [---] engagements
"Okay this is insane. WebGPU-accelerated semantic video tracking powered by DINOv3 and Transformers.js 🤯 This will revolutionize AI-powered video editors. which can now run 100% locally in your browser no server inference required (costs $0) 😍 Who's building this"
X Link 2025-08-22T14:05Z 15.1K followers, 43.7K engagements
"@LiquidAI_ @KarnikShreyas It always amazes me to see what the community is capable of building with this stuff. 🤗 Great work @KarnikShreyas I tested it by analyzing the config of Grok-2:"
X Link 2025-08-27T16:55Z 13.8K followers, [---] engagements
"NEW: Google releases EmbeddingGemma a state-of-the-art multilingual embedding model perfect for on-device use cases At only 308M params the model can run 100% locally in your browser 🤯 Explore your documents in an interactive 3D universe with our demo: "The Semantic Galaxy""
X Link 2025-09-04T16:21Z 15.1K followers, 23.9K engagements
"IBM just released Granite [---] their latest series of small language models These models excel at agentic workflows (tool calling) document analysis RAG and more. 🚀 The "Micro" (3.4B) model can even run 100% locally in your browser on WebGPU powered by 🤗 Transformers.js"
X Link 2025-10-02T16:16Z 15.1K followers, 84.7K engagements
"🗂 Model collection: 🔗 WebGPU demo + source code: https://huggingface.co/spaces/ibm-granite/Granite-4.0-WebGPU https://huggingface.co/collections/ibm-granite/granite-40-language-models-6811a18b820ef362d9e5a82c https://huggingface.co/spaces/ibm-granite/Granite-4.0-WebGPU https://huggingface.co/collections/ibm-granite/granite-40-language-models-6811a18b820ef362d9e5a82c"
X Link 2025-10-02T16:16Z 14.1K followers, [----] engagements
"🔗 Demo + source code: https://huggingface.co/spaces/ibm-granite/granite-docling-258M-WebGPU https://huggingface.co/spaces/ibm-granite/granite-docling-258M-WebGPU"
X Link 2025-10-07T19:46Z 14.2K followers, [----] engagements
"Chrome's new window.ai feature is going to change the web forever 🤯 It allows you to run Gemini Nano a powerful 3.25B parameter LLM 100% locally in your browser We've also added experimental support to 🤗 Transformers.js making it super easy to use 😍 Check it out 👇"
X Link 2024-07-08T16:54Z 15.1K followers, 583.8K engagements
"First project of 2025: Vision Transformer Explorer I built a web app to interactively explore the self-attention maps produced by ViTs. This explains what the model is focusing on when making predictions and provides insights into its inner workings 🤯 Try it out yourself 👇"
X Link 2025-01-01T15:03Z 12.5K followers, [----] engagements
""DeepSeek-R1-Distill-Qwen-1.5B outperforms GPT-4o and Claude-3.5-Sonnet on math benchmarks with 28.9% on AIME and 83.9% on MATH." We now have open-source reasoning models that outperform GPT-4o and run 100% locally in your browser on WebGPU 🤯 [----] is off to a wild start 🔥"
X Link 2025-01-21T20:47Z 15.1K followers, 174.6K engagements
"DeepSeek is on FIRE 🔥 They just released Janus Pro: a multimodal LLM capable of visual understanding and image generation 🤯 The 1B model can even run in your browser on WebGPU powered by 🤗 Transformers.js This is the easiest way to run it locally: just visit a website"
X Link 2025-01-27T23:39Z 15.1K followers, 244.6K engagements
"We did it. Kokoro TTS (v1.0) can now run 100% locally in your browser w/ WebGPU acceleration. Real-time text-to-speech without a server. ⚡ Generate [--] seconds of speech in [--] second for $0. What will you build 🔥 If someone wants a gig here is what I'd like: - convert to WebGPU - implement sentence splitting - register some nifty url with just a upload text or past from clipboard button (like https://t.co/b8lbYALU5u or whatever) - that instantly starts an audio player - save prior If someone wants a gig here is what I'd like: - convert to WebGPU - implement sentence splitting - register some"
X Link 2025-02-07T15:16Z 15.1K followers, 69.8K engagements
"Introducing Kokoro Web: ML-powered speech synthesis directly in your browser 🚀 It's 100% free and supports unlimited generations. Choose from [--] unique voices offering natural and expressive speech for any application. Now w/ streaming & WebGPU acceleration Check it out 👇"
X Link 2025-02-14T17:35Z 15.1K followers, 37.4K engagements
"NEW: Real-time conversational AI models can now run 100% locally in your browser 🤯 🔐 Privacy by design (no data leaves your device) 💰 Completely free. forever 📦 Zero installation required just visit a website ⚡ Blazingly-fast WebGPU-accelerated inference Try it out 👇"
X Link 2025-06-04T18:30Z 15.1K followers, 14.2K engagements
"Introducing Granite Docling WebGPU 🐣 State-of-the-art document parsing 100% locally in your browser 🤯 🔐 No data sent to a server (private & secure) 💰 Completely free. forever 🔂 Docling ecosystem enables conversion to HTML Markdown JSON and more Try out the demo 👇"
X Link 2025-10-07T19:44Z 15.1K followers, 57.1K engagements
"BOOM 💥 Today I added WebGPU support for @karpathy's nanochat models meaning they can run 100% locally in your browser (no server) The d32 version runs at over [--] tps on my M4 Max 🚀 Pretty wild that you can now deploy AI applications using just a single index.html file 😅 Excited to release new repo: nanochat (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining nanochat is a minimal from scratch full-stack training/inference pipeline of a simple ChatGPT clone in a single https://t.co/LLhbLCoZFt Excited to release new repo:"
X Link 2025-10-20T22:23Z 15.1K followers, 96.4K engagements
"IBM just released Granite-4.0 Nano their smallest LLMs ever (300M & 1B) 😍 The models demonstrate remarkable instruction following and tool calling capabilities and can even run locally in-browser This means they can interact with websites and call browser APIs for you 🤯"
X Link 2025-10-28T17:06Z 15.1K followers, 33.1K engagements
"I can't wait to see what the community builds with it 🤗 Links: 📄 Blog post: 🚀 Demo (+ source code): https://huggingface.co/spaces/ibm-granite/Granite-4.0-Nano-WebGPU https://huggingface.co/blog/ibm-granite/granite-4-nano https://huggingface.co/spaces/ibm-granite/Granite-4.0-Nano-WebGPU https://huggingface.co/blog/ibm-granite/granite-4-nano"
X Link 2025-10-28T17:06Z 14.4K followers, [----] engagements
"Meta Superintelligence Labs just released SAM 3: Segment Anything with Concepts. A unified model that detects segments and tracks objects in images or videos using concept prompts (text or images) 🤯 Model weights (+ demo) now out on 🤗 Hugging Face"
X Link 2025-11-19T16:20Z 14.5K followers, [----] engagements
"SAM [--] can even run 100% locally in your browser on WebGPU enabling real-time image segmentation ⚡ Original model: ONNX model: WebGPU demo: https://huggingface.co/spaces/webml-community/SAM3-Tracker-WebGPU https://huggingface.co/onnx-community/sam3-tracker-ONNX https://huggingface.co/facebook/sam3 https://huggingface.co/spaces/webml-community/SAM3-Tracker-WebGPU https://huggingface.co/onnx-community/sam3-tracker-ONNX https://huggingface.co/facebook/sam3"
X Link 2025-11-19T16:55Z 14.5K followers, [----] engagements
"Brilliant 🤩 This is a project Ive been wanting to see for so long Congrats on getting in done 🤗 If you want to go a step further: aligning the API exactly with PyTorch so that students can learn/practice how to build neural networks directly in-browser would be an invaluable teaching resource. Behind the scenes of course would need to be some transpilation magic to add async/awaits (to make JS happy) 😁 Excited to see the progression of this project"
X Link 2025-11-21T14:54Z 14.4K followers, [---] engagements
"NEW: @MistralAI releases Mistral [--] a family of multimodal models including three start-of-the-art dense models (3B 8B and 14B) and Mistral Large [--] (675B 41B active). All Apache [---] 🤗 Surprisingly the 3B is small enough to run 100% locally in your browser on WebGPU 🤯 https://twitter.com/i/web/status/1995879338583945635 https://twitter.com/i/web/status/1995879338583945635"
X Link 2025-12-02T15:34Z 14.9K followers, 224.9K engagements
"Link to collections: - Mistral Large [--] (675B 41B active): - Dense models (3B 8B and 14B): Try out the demo yourself 👇 https://huggingface.co/spaces/mistralai/Ministral_3B_WebGPU https://huggingface.co/collections/mistralai/ministral-3 https://huggingface.co/collections/mistralai/mistral-large-3 https://huggingface.co/spaces/mistralai/Ministral_3B_WebGPU https://huggingface.co/collections/mistralai/ministral-3 https://huggingface.co/collections/mistralai/mistral-large-3"
X Link 2025-12-02T15:34Z 14.9K followers, [----] engagements
"Ultra-fast expressive voice cloning is now possible with Chatterbox Turbo 🤯 Use paralinguistic tags like chuckle or gasp to enhance the realism of your generations and clone voices with just [--] seconds of reference audio. Now available on Hugging Face 🤗"
X Link 2025-12-15T20:01Z 14.9K followers, 10.9K engagements
"NEW: Google releases FunctionGemma a lightweight (270M) open foundation model built for creating specialized function calling models 🤯 To test it out I built a small game: use natural language to solve fun physics simulation puzzles running 100% locally in your browser 🕹"
X Link 2025-12-18T17:19Z 14.9K followers, 57.1K engagements
"microgpt.js: the most atomic way to train and inference a GPT in pure dependency-free JavaScript. This repository covers the complete algorithm. Everything else is just efficiency. Links: - Demo: - GitHub: https://github.com/xenova/microgpt.js https://huggingface.co/spaces/webml-community/microgpt.js https://github.com/xenova/microgpt.js https://huggingface.co/spaces/webml-community/microgpt.js"
X Link 2026-02-13T22:30Z 15.1K followers, [----] engagements
"Introducing Voxtral WebGPU: State-of-the-art audio transcription directly in your browser 🤯 🗣 Transcribe videos meeting notes songs and more 🔐 Runs on-device meaning no data is sent to a server 🌎 Multilingual (8 languages) 🤗 Completely free (forever) & open source https://twitter.com/i/web/status/1948404315934077354 https://twitter.com/i/web/status/1948404315934077354"
X Link 2025-07-24T15:26Z 15.1K followers, 20.6K engagements
"🫡 Good luck to your Mac It should work provided you have enough VRAM. 🤞 https://twitter.com/i/web/status/2022358009616204151 https://twitter.com/i/web/status/2022358009616204151"
X Link 2026-02-13T17:11Z 15.1K followers, [---] engagements
"After nearly a year of development 🤗 Transformers.js v4 Preview is finally out on npm npm i @huggingface/transformers@next Build WebGPU-accelerated AI applications that run everywhere: browsers Node.js Bun Deno Electron and more. See what's new in our blog post 👇"
X Link 2026-02-09T18:00Z 15.1K followers, 25.4K engagements
"Behold. GPT-OSS (20B) running 100% locally in your browser on WebGPU. This shouldn't be possible but with Transformers.js v4 and ONNX Runtime Web it is A new class of AI apps is emerging. Zero-install infinite distribution. Simply visit a website and run models locally"
X Link 2026-02-13T16:48Z 15.1K followers, 304.4K engagements
"Inspired by @karpathy's microgpt I built microgpt.js: a JavaScript port that runs entirely in your browser It's an exact numerical implementation so the randomness and outputs match bit-for-bit Try it out yourself and train your own GPT by simply opening a webpage 👇 New art project. Train and inference GPT in [---] lines of pure dependency-free Python. This is the full algorithmic content of what is needed. Everything else is just for efficiency. I cannot simplify this any further. https://t.co/HmiRrQugnP New art project. Train and inference GPT in [---] lines of pure dependency-free Python."
X Link 2026-02-13T22:30Z 15.1K followers, 48.4K engagements
"I get over [--] tokens/second on my M4 Max. what about you Try it out yourself and let me know 👇 https://huggingface.co/spaces/webml-community/GPT-OSS-WebGPU https://huggingface.co/spaces/webml-community/GPT-OSS-WebGPU"
X Link 2026-02-13T16:48Z 15.1K followers, [----] engagements
"Learn more about what else is now possible with 🤗 Transformers.js v4 https://x.com/xenovacom/status/2020920634944155800 After nearly a year of development 🤗 Transformers.js v4 Preview is finally out on npm npm i @huggingface/transformers@next Build WebGPU-accelerated AI applications that run everywhere: browsers Node.js Bun Deno Electron and more. See what's new in our blog post 👇 https://t.co/dX4m5GMkbx https://x.com/xenovacom/status/2020920634944155800 After nearly a year of development 🤗 Transformers.js v4 Preview is finally out on npm npm i @huggingface/transformers@next Build"
X Link 2026-02-14T00:22Z 15.1K followers, [----] engagements
"If you're not following what's happening with small specialized language models you're not paying close enough attention. 4B models can be trained to solve Olympiad-level math problems just as well as models 30x larger (GPT-OSS-120B). They can even run in-browser on WebGPU"
X Link 2026-02-16T15:45Z 15.1K followers, [----] engagements
"QED-Nano WebGPU: Frontier-level math AI running 100% locally in your browser. Powered by Transformers.js v4 - Blog post: - Model: - Demo (& source code): https://huggingface.co/spaces/webml-community/QED-Nano-WebGPU https://huggingface.co/lm-provers/QED-Nano https://huggingface.co/spaces/lm-provers/qed-nano-blogpost https://huggingface.co/spaces/webml-community/QED-Nano-WebGPU https://huggingface.co/lm-provers/QED-Nano https://huggingface.co/spaces/lm-provers/qed-nano-blogpost"
X Link 2026-02-16T15:45Z 15.1K followers, [---] engagements
"After nearly a year of development 🤗 Transformers.js v4 Preview is finally out on npm npm i @huggingface/transformers@next Build WebGPU-accelerated AI applications that run everywhere: browsers Node.js Bun Deno Electron and more. See what's new in our blog post 👇"
X Link 2026-02-09T18:00Z 15.1K followers, 25.4K engagements
"NEW: Apple releases FastVLM and MobileCLIP2 on Hugging Face 🤗 The models are up to 85x faster and 3.4x smaller than previous work enabling real-time VLM applications 🤯 It can even do live video captioning 100% locally in your browser (zero install). Huge for accessibility"
X Link 2025-08-29T15:42Z 15.1K followers, 224.7K engagements
"It's finally possible: real-time in-browser speech recognition with OpenAI Whisper 🤯 The model runs fully on-device using Transformers.js and ONNX Runtime Web and supports multilingual transcription across [---] different languages 🔥 Check out the demo (+ source code) 👇"
X Link 2024-06-07T16:05Z 15.1K followers, 266.9K engagements
"There has been a huge debate recently about the best approach for image background removal. Here's my attempt: - In-browser inference w/ 🤗 Transformers.js - WebGPU accelerated (fast) - Costs $0 (no image hosting or server processing) - No data leaves your device (privacy) It works. It's fast. It's super high quality. It's cheap. And it scales. Available for purchase soon https://t.co/5If4T3Inj3 It works. It's fast. It's super high quality. It's cheap. And it scales. Available for purchase soon https://t.co/5If4T3Inj3"
X Link 2024-08-26T17:07Z 15.1K followers, 422.3K engagements
"It works. It's fast. It's super high quality. It's cheap. And it scales. Available for purchase soon"
X Link 2024-08-26T10:58Z 282.1K followers, 823.4K engagements
""DeepSeek-R1-Distill-Qwen-1.5B outperforms GPT-4o and Claude-3.5-Sonnet on math benchmarks with 28.9% on AIME and 83.9% on MATH." We now have open-source reasoning models that outperform GPT-4o and run 100% locally in your browser on WebGPU 🤯 [----] is off to a wild start 🔥"
X Link 2025-01-21T20:47Z 15.1K followers, 174.6K engagements
"DeepSeek is on FIRE 🔥 They just released Janus Pro: a multimodal LLM capable of visual understanding and image generation 🤯 The 1B model can even run in your browser on WebGPU powered by 🤗 Transformers.js This is the easiest way to run it locally: just visit a website"
X Link 2025-01-27T23:39Z 15.1K followers, 244.6K engagements
"Run OpenAI's new Whisper Turbo model 100% locally in your browser with Transformers.js ⚡ Transcribe [--] minutes of audio in [--] seconds 🤯 Demo + source code 👇"
X Link 2024-10-01T11:37Z 15.1K followers, 141.7K engagements
"Introducing Whisper Diarization: Multilingual speech recognition with word-level timestamps and speaker segmentation running 100% locally in your browser thanks to 🤗 Transformers.js Tested on this iconic Lettermen interview w/ Grace Hopper from [----] Demo (+ source code) 👇"
X Link 2024-07-22T16:47Z 15.1K followers, 94.9K engagements
"Introducing Whisper Web: ML-powered speech recognition directly in your browser 🚀 This comes with the release of Transformers.js v2.2.0 which now supports multilingual transcription and translation for over [---] different languages 🤯 Check it out: https://huggingface.co/spaces/Xenova/whisper-web https://huggingface.co/spaces/Xenova/whisper-web"
X Link 2023-06-09T16:42Z 15.1K followers, 1.1M engagements
"I'm excited to announce that Transformers.js V3 is finally available on NPM 🔥 State-of-the-art Machine Learning for the web now with WebGPU support 🤯⚡ Install it from NPM with: @/ or via CDN (example below) 👇"
X Link 2024-08-12T16:00Z 15.1K followers, 89.4K engagements
"Chrome's new window.ai feature is going to change the web forever 🤯 It allows you to run Gemini Nano a powerful 3.25B parameter LLM 100% locally in your browser We've also added experimental support to 🤗 Transformers.js making it super easy to use 😍 Check it out 👇"
X Link 2024-07-08T16:54Z 15.1K followers, 583.8K engagements
"Meta's Segment Anything Model (SAM) can now run in your browser w/ WebGPU (+ fp16) meaning up to 8x faster image encoding (10s 1.25s) 🤯⚡ Video is not sped up Everything runs 100% locally thanks to 🤗 Transformers.js and onnxruntime-web 🔗 Demo: https://hf.co/spaces/Xenova/segment-anything-webgpu https://hf.co/spaces/Xenova/segment-anything-webgpu"
X Link 2024-04-21T00:44Z 15.1K followers, 122.4K engagements
"IBM just released Granite [---] their latest series of small language models These models excel at agentic workflows (tool calling) document analysis RAG and more. 🚀 The "Micro" (3.4B) model can even run 100% locally in your browser on WebGPU powered by 🤗 Transformers.js"
X Link 2025-10-02T16:16Z 15.1K followers, 84.7K engagements
"Introducing Whisper Timestamped: Multilingual speech recognition with word-level timestamps running 100% locally in your browser thanks to 🤗 Transformers.js This unlocks a world of possibilities for in-browser video editing 🤯 What will you build 😍 Demo (+ source code) 👇"
X Link 2024-07-10T16:00Z 15.1K followers, 108.2K engagements
"Florence-2 the new vision foundation model by Microsoft can now run 100% locally in your browser on WebGPU thanks to Transformers.js 🤗🤯 It supports tasks like image captioning optical character recognition object detection and many more 😍 WOW Demo (+ source code) 👇"
X Link 2024-06-26T15:42Z 15.1K followers, 90.6K engagements
"Is this the future of AI browser agents 👀 WebGPU-accelerated reasoning LLMs are now supported in Transformers.js 🤯 Here's MiniThinky-v2 (1B) running 100% locally in the browser at [--] tps (no API calls) I can't wait to see what you build with it Demo + source code in 🧵👇"
X Link 2025-01-10T12:08Z 15.1K followers, 83.5K engagements
"Introducing Whisper WebGPU: Blazingly-fast ML-powered speech recognition directly in your browser 🚀 It supports multilingual transcription and translation across [---] languages 🤯 The model runs locally meaning no data leaves your device 😍 Check it out (+ source code) 👇"
X Link 2024-06-09T17:38Z 15.1K followers, 95.3K engagements
"I seriously cannot believe this is a 0.6B LLM 🤯 @Alibaba_Qwen just released Qwen3 a series of hybrid reasoning models that allow you to control how much "thinking" the model does for a given task. They can even run locally in your browser on WebGPU with 🤗 Transformers.js"
X Link 2025-04-29T01:46Z 15.1K followers, 76.4K engagements
"After more than a year of development we're excited to announce the release of 🤗 Transformers.js v3 ⚡ WebGPU support (up to 100x faster than WASM) 🔢 New quantization formats (dtypes) 🏛 [---] supported architectures in total 📂 [--] new example projects and templates 🤖 Over [----] pre-converted models 🌐 Node.js (ESM + CJS) Deno and Bun compatibility 🏡 A new home on GitHub and NPM Get started with npm i @huggingface/transformers. Learn more in the blog post below 👇"
X Link 2024-10-22T15:02Z 15.1K followers, 44.7K engagements
"OmniParser the new screen parsing tool from Microsoft (and #1 trending model on @huggingface) can now run 100% locally in your browser with Transformers.js 🤯 Who's going to be the first to turn this into a browser extension 👀 Endless possibilities Demo & code below 👇"
X Link 2024-11-01T07:29Z 15.1K followers, 66.2K engagements
"Introducing Phi-3 WebGPU a private and powerful AI chatbot that runs locally in your browser powered by 🤗 Transformers.js and onnxruntime-web 🔒 On-device inference: no data sent to a server ⚡ WebGPU-accelerated ( [--] t/s) 📥 Model downloaded once and cached Try it out 👇"
X Link 2024-05-08T12:00Z 15.1K followers, 119K engagements
"Introducing Kokoro.js a new JavaScript library for running Kokoro TTS an [--] million parameter text-to-speech model 100% locally in the browser w/ WASM. Powered by 🤗 Transformers.js. WebGPU support coming soon 👉 npm i kokoro-js 👈 Link to demo (+ sample code) in 🧵"
X Link 2025-01-16T14:52Z 15.1K followers, 42.6K engagements
"We just updated our in-browser Background Removal demo to use WebGPU and it's now 50x faster 🤯 [--] seconds down to 180ms ⚡ Powered by @bria_ai_'s RMBG-v1.4 model and 🤗 Transformers.js . and yes the video is in real time 🤯"
X Link 2024-03-09T21:16Z 15.1K followers, 76.6K engagements
"BOOM 💥 Today I added WebGPU support for @karpathy's nanochat models meaning they can run 100% locally in your browser (no server) The d32 version runs at over [--] tps on my M4 Max 🚀 Pretty wild that you can now deploy AI applications using just a single index.html file 😅 Excited to release new repo: nanochat (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining nanochat is a minimal from scratch full-stack training/inference pipeline of a simple ChatGPT clone in a single https://t.co/LLhbLCoZFt Excited to release new repo:"
X Link 2025-10-20T22:23Z 15.1K followers, 96.4K engagements
"Excited to release new repo: nanochat (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining nanochat is a minimal from scratch full-stack training/inference pipeline of a simple ChatGPT clone in a single dependency-minimal codebase. You boot up a cloud GPU box run a single script and in as little as [--] hours later you can talk to your own LLM in a ChatGPT-like web UI. It weighs [----] lines of imo quite clean code to: - Train the tokenizer using a new Rust implementation - Pretrain a Transformer LLM on FineWeb evaluate CORE score"
X Link 2025-10-13T15:16Z 1.8M followers, 5.8M engagements
"Introducing Granite Docling WebGPU 🐣 State-of-the-art document parsing 100% locally in your browser 🤯 🔐 No data sent to a server (private & secure) 💰 Completely free. forever 🔂 Docling ecosystem enables conversion to HTML Markdown JSON and more Try out the demo 👇"
X Link 2025-10-07T19:44Z 15.1K followers, 57.1K engagements
"We did it. Kokoro TTS (v1.0) can now run 100% locally in your browser w/ WebGPU acceleration. Real-time text-to-speech without a server. ⚡ Generate [--] seconds of speech in [--] second for $0. What will you build 🔥 If someone wants a gig here is what I'd like: - convert to WebGPU - implement sentence splitting - register some nifty url with just a upload text or past from clipboard button (like https://t.co/b8lbYALU5u or whatever) - that instantly starts an audio player - save prior If someone wants a gig here is what I'd like: - convert to WebGPU - implement sentence splitting - register some"
X Link 2025-02-07T15:16Z 15.1K followers, 69.8K engagements
"If someone wants a gig here is what I'd like: - convert to WebGPU - implement sentence splitting - register some nifty url with just a upload text or past from clipboard button (like or whatever) - that instantly starts an audio player - save prior uploads and play progress in local storage - consider for downloading weights if not present no server all local essentially zero running cost public good. https://webtorrent.io/ http://read4.me https://webtorrent.io/ http://read4.me"
X Link 2025-01-13T16:50Z 448.9K followers, 408.9K engagements
"New @karpathy video just dropped 😍🔥 After watching if you want to learn more about how different models (e.g. GPT4 Llama T5 BERT) tokenize text check out "The Tokenizer Playground": a web-app I built a few months ago with 🤗 Transformers.js 🔗 https://huggingface.co/spaces/Xenova/the-tokenizer-playground New (2h13m 😅) lecture: "Let's build the GPT Tokenizer" Tokenizers are a completely separate stage of the LLM pipeline: they have their own training set training algorithm (Byte Pair Encoding) and after training implement two functions: encode() from strings to tokens and"
X Link 2024-02-20T21:18Z 15.1K followers, 77.3K engagements
"New (2h13m 😅) lecture: "Let's build the GPT Tokenizer" Tokenizers are a completely separate stage of the LLM pipeline: they have their own training set training algorithm (Byte Pair Encoding) and after training implement two functions: encode() from strings to tokens and decode() back from tokens to strings. In this lecture we build from scratch the Tokenizer used in the GPT series from OpenAI"
X Link 2024-02-20T17:40Z 1.8M followers, 1.7M engagements
"Woah. This is the fastest Transformers.js has ever been 🤯 Run a 1.7B LLM 100% locally in your browser at over [---] tokens per second 🚀 No server required. ⚡ WebGPU-accelerated in-browser inference 📦 Optimized ONNX exports w/ GQA 💬 Multi-round conversations w/ KV caching"
X Link 2025-04-22T00:04Z 15.1K followers, 55.2K engagements
"DINOv3 is revolutionary: a new state-of-the-art vision backbone trained to produce rich dense image features. I loved their demo video so much that I decided to re-create their visualization tool. Everything runs 100% in-browser with 🤗 Transformers.js Demo + source code 👇 Introducing DINOv3: a state-of-the-art computer vision model trained with self-supervised learning (SSL) that produces powerful high-resolution image features. For the first time a single frozen vision backbone outperforms specialized solutions on multiple long-standing dense https://t.co/nwS3zFCaaN Introducing DINOv3: a"
X Link 2025-08-15T21:49Z 15.1K followers, 75.7K engagements
"Introducing DINOv3: a state-of-the-art computer vision model trained with self-supervised learning (SSL) that produces powerful high-resolution image features. For the first time a single frozen vision backbone outperforms specialized solutions on multiple long-standing dense prediction tasks. Learn more about DINOv3 here: https://ai.meta.com/blog/dinov3-self-supervised-vision-model/utm_source=twitter&utm_medium=organic_social&utm_content=video&utm_campaign=dinov3"
X Link 2025-08-14T16:19Z 750.5K followers, 902.1K engagements
"Introducing Distil-Whisper Web: 49% smaller 4.2x faster Whisper directly in your browser 🚀 Here is a side-by-side comparison with OpenAI's original version 🤯"
X Link 2023-11-03T15:20Z 15.1K followers, 273.8K engagements
"Depth Anything V2 just released enabling real-time depth estimation directly in your browser with 🤗 Transformers.js and WebGPU acceleration ⚡ The smallest model is only 50MB (@ fp16) making it perfect for on-device usage 😍 Check out the demo (+ source code) 👇"
X Link 2024-06-14T17:45Z 15.1K followers, 42.6K engagements
"WebGPU is the future 😍🔥 Transformers.js can now perform real-time background removal powered by MODNet ⚡ Development for Transformers.js v3 (which adds full WebGPU support) is well underway and we're excited to continue sharing updates and demos Try it out yourself 👇"
X Link 2024-03-13T22:00Z 15.1K followers, 47.4K engagements
"Okay this is insane. WebGPU-accelerated semantic video tracking powered by DINOv3 and Transformers.js 🤯 This will revolutionize AI-powered video editors. which can now run 100% locally in your browser no server inference required (costs $0) 😍 Who's building this"
X Link 2025-08-22T14:05Z 15.1K followers, 43.7K engagements
"IBM just released Granite-4.0 Nano their smallest LLMs ever (300M & 1B) 😍 The models demonstrate remarkable instruction following and tool calling capabilities and can even run locally in-browser This means they can interact with websites and call browser APIs for you 🤯"
X Link 2025-10-28T17:06Z 15.1K followers, 33.1K engagements
"I know it just released but I don't see many people talking about the Phi-3 tokenizer 👀 Here's the full list of added special tokens. what do you notice 🤯 assistant step function_output tag function_call system end raw continue user function_list calc code /code summary resource assistant_mask start message fim_prefix fim_middle fim_suffix meta_start ipynb_marker diff_marker ghissue ghreview disc_start disc_sep disc_threadquery /query data /data sys /sys inst /inst"
X Link 2024-04-23T15:52Z 15.1K followers, 100.1K engagements
"Llama [---] running 100% locally in your browser on WebGPU 🦙 Up to [--] tokens per second ⚡ Powered by 🤗 Transformers.js and ONNX Runtime Web. No installation required. just visit a website Check out the demo and source code below 👇"
X Link 2024-09-30T14:56Z 15.1K followers, 35.2K engagements
"Introducing Kokoro Web: ML-powered speech synthesis directly in your browser 🚀 It's 100% free and supports unlimited generations. Choose from [--] unique voices offering natural and expressive speech for any application. Now w/ streaming & WebGPU acceleration Check it out 👇"
X Link 2025-02-14T17:35Z 15.1K followers, 37.4K engagements
"Introducing Moonshine Web: real-time speech recognition running 100% locally in your browser 🚀 Faster and more accurate than Whisper 🔒 Privacy-focused (no data leaves your device) ⚡ WebGPU accelerated (w/ WASM fallback) 🔥 Powered by ONNX Runtime Web and Transformers.js"
X Link 2024-12-18T16:42Z 15.1K followers, 28.8K engagements
"Massive size reductions for Transformers.js models: - Whisper-tiny decoder: 50MB 30MB (-40%) - NLLB decoder: 732MB 476MB (-35%) - bloom: 819MB 562MB (-31%) - gpt2: 167MB 127MB (-24%) - bert-base: 134MB 111MB (-17%) . all this with ZERO drop in performance 🤯"
X Link 2023-09-04T17:00Z 15.1K followers, 108K engagements
"YOLOv9 just released and now it's compatible with 🤗 Transformers.js That's right. near real-time object detection running locally in your browser: no server required 🤯 Try it out yourself 👇 🔗 Demo: https://hf.co/spaces/Xenova/yolov9-web https://hf.co/spaces/Xenova/yolov9-web"
X Link 2024-02-23T18:31Z 15.1K followers, 71.1K engagements
"When do you really need to use a vector database 🤔 To try answer that question I recreated my semantic image search application to run 100% in-browser with Transformers.js (no server). After loading the model and database it only takes 50ms to compute text embeddings and perform similarity search across 25k images. No vector DB just plain old JavaScript. I think @karpathy was onto something. 👀 What do you think 🔗 Demo site: Source code: https://github.com/xenova/transformers.js/tree/main/examples/semantic-image-search-client"
X Link 2023-09-23T00:57Z 15.1K followers, 186.8K engagements
"@sinclanich np.array people keep reaching for much fancier things way too fast these days"
X Link 2023-04-15T23:01Z 1.8M followers, 663.3K engagements
"Introducing SmolLM: a new SOTA series of 135M 360M and 1.7B models perfect for on-device deployment 🔥 We also uploaded ONNX weights for the models meaning they can run locally in your browser with 🤗 Transformers.js and WebGPU acceleration ⚡ Try it out (+ blog post) 👇"
X Link 2024-07-16T17:03Z 15.1K followers, 34K engagements
"Transformers.js v2.0 is finally here 🔥 Run @huggingface transformers directly in your browser with no need for a server Some of the new features include: 🛠 Complete ES6 rewrite 📄 Documentation + examples 🤗 Improved Hugging Face Hub integration https://github.com/xenova/transformers.js https://github.com/xenova/transformers.js"
X Link 2023-05-17T17:54Z 15.1K followers, 67K engagements
"Jina CLIP v1 just released: a new state-of-the-art multimodal embedding model that outperforms OpenAI CLIP in text-image retrieval 😍 We also contributed ONNX weights so it's now compatible with 🤗 Transformers.js v3 and runs with WebGPU acceleration ⚡ Try out the demo 👇"
X Link 2024-06-05T11:50Z 15.1K followers, 45.3K engagements
"Introducing TTS WebGPU: The first ever text-to-speech web app built with WebGPU acceleration 🔥 High-quality and natural speech generation that runs 100% locally in your browser powered by OuteTTS and Transformers.js.🤗 Try it out yourself Demo + source code below 👇"
X Link 2024-12-08T19:32Z 15.1K followers, 21.3K engagements
"WOW 🤯 DINOv3 can run locally on your phone. from the browser This unlocks endless possibilities for AI-powered web apps. 🤏 Model is tiny (only 15MB at 4-bit quantization) 🧠 Delivers powerful high-resolution image features ✨ Works completely offline Try it yourself 👇"
X Link 2025-08-16T17:04Z 15.1K followers, 40K engagements
"I can't believe this. Phi-3.5-mini (3.8B) running in-browser at [--] tokens/second on WebGPU w/ Transformers.js and ONNX Runtime Web 🤯 Since everything runs 100% locally no messages are sent to a server a huge win for privacy 🔒 Check out the demo + source code below 👇"
X Link 2024-08-23T14:40Z 15.1K followers, 22.3K engagements
Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing