@QuixiAI Eric Hartford

Eric Hartford posts on X about model, this is, ai, to the the most. They currently have [------] followers and [---] posts still getting attention that total [------] engagements in the last [--] hours.

Engagements: [------] #

• [--] Week [------] +210%
• [--] Month [-------] -68%
• [--] Months [---------] +18%
• [--] Year [---------] +55%

Mentions: [--] #

• [--] Week [--] +72%
• [--] Month [--] +22%
• [--] Months [---] +2.20%
• [--] Year [---] +0.19%

Followers: [------] #

• [--] Week [------] +0.24%
• [--] Month [------] +0.14%
• [--] Months [------] +4.60%
• [--] Year [------] +13%

CreatorRank: [-------] #

Social Influence

Social category influence technology brands 16.24% stocks 8.55% finance 2.56% vc firms 2.56% countries 1.71% celebrities 1.71% cryptocurrencies 0.85%

Social topic influence model 14.53%, this is 10.26%, ai 7.69%, to the 5.98%, sponsor 5.98%, in the 5.98%, if you 5.98%, sama 4.27%, open ai 4.27%, chat 4.27%

Top accounts mentioned or mentioned by @fernandonetoai @latkins @huggingface @crusoecloud @sama @openai @crusoeenergy @aiatmeta @liquidai @hotaisle @mistralai @adamholtererer @googledeepmind @msftresearch @ibm @a16z @abacusai @tensorwave @deepseekai @minimaxai

Top assets mentioned IBM (IBM) Microsoft Corp. (MSFT) Alphabet Inc Class A (GOOGL) Flex Ltd. Ordinary Shares (FLEX)

Top Social Posts

Top posts by engagements in the last [--] hours

"Guys don't buy a Mac mini for this You can run it in a docker container Or an ec2 instance Or a raspberry pi Or any old laptop you can leave plugged in There's literally no reason to buy a whole desktop computer just to run it spent $799 on a Mac Mini just to realize Clawdbot is useless now this thing is collecting dust https://t.co/nVWdJ9RQMf spent $799 on a Mac Mini just to realize Clawdbot is useless now this thing is collecting dust https://t.co/nVWdJ9RQMf"
X Link 2026-01-30T17:35Z 18K followers, [----] engagements

"Running Kimi K2.5 with @vllm_project on my mi300x - It turns out the weights aren't China-censored"
X Link 2026-01-31T00:21Z 18K followers, [----] engagements

""Capturing CUDA graphs" - We are chasing them That's why it takes a long time"
X Link 2026-02-01T07:23Z 18K followers, [---] engagements

"I know it's probably a great model I just wish they didnt cherry pick their benchmarks so much. Like where is MMLU HLE ARC AGI. Too bad @Huggingface shut down their leaderboard and nobody else has stepped up. We also learned from @AIatMeta that we can't just take the model publishers word for it. Step-3.5-Flash by @StepFun_ai looks promising https://t.co/wFxLozaOjY Step-3.5-Flash by @StepFun_ai looks promising https://t.co/wFxLozaOjY"
X Link 2026-02-02T09:15Z 18K followers, [----] engagements

"Today gpt5.2 codex v high tried and failed for [--] minutes while Claude [---] Opus fixed me in [--] minutes. System troubleshooting"
X Link 2026-02-05T23:20Z 18K followers, [----] engagements

"@Alibaba_Qwen's biggest flex with Qwen3-Coder-Next (80b) is that the models they compared themselves against are 3x-8x the size. (230b 358b 685b) Amazing"
X Link 2026-02-08T02:22Z 18K followers, [----] engagements

"GPT-5.3 Codex may be better at coding than Claude Opus [---] - as the benchmarks say - but Claude Code with Opus [---] is better and faster at getting work done than Codex CLI with GPT-5.3 Codex. 2x-3x faster. (measured in how much work I get done per hour)"
X Link 2026-02-08T04:17Z 18K followers, [----] engagements

"Qwen3.5 architecture is out: A vision language model hybrid SSM-Transformer using Gated DeltaNet linear attention mixed with standard attention interleaved MRoPE and shared+routed MoE experts. https://github.com/huggingface/transformers/pull/43830/ https://github.com/huggingface/transformers/pull/43830/"
X Link 2026-02-10T06:31Z 18.1K followers, [----] engagements

"@RayFernando1337 @exolabs Quad-3090 is faster and cheaper than dgx spark it's criminal that dgx spark gets so much attention when its so underpowered for the price"
X Link 2026-02-10T16:17Z 18K followers, [--] engagements

"no. bitnet is not a quant. you have to train a bitnet. that is why there are only the [--] that Microsoft published - 1b and 2b. Also Microsoft did not share their training code so nobody but Microsoft can train them (unless they reimplement from scratch by looking at the Bitnet paper) https://twitter.com/i/web/status/2021344540674621771 https://twitter.com/i/web/status/2021344540674621771"
X Link 2026-02-10T22:04Z 18K followers, [--] engagements

"@casper_hansen_ How is that cheating"
X Link 2026-02-11T02:44Z 18K followers, [---] engagements

"Congrats @Zai_org on GLM-5 Love the permissive MIT license (vs K2.5's modified MIT). Haven't chatted with it yet so no vibes but from the numbers I'm not compelled to switch from @Kimi_Moonshot K2.5: Similar evals but GLM-5's are at bf16 while K2.5's are at int4 - GLM-5 will lose ground once quantized K2.5 has vision (image + video) GLM-5 doesn't GLM-5 is smaller so slightly cheaper to host once quantized That said I appreciate @Zai_org's honesty with their evals and the truly permissive license. https://twitter.com/i/web/status/2021651135615184988"
X Link 2026-02-11T18:22Z 18K followers, [----] engagements

"I've been saying this for six months. You need to get ready for this. https://t.co/ivXRKXJvQg https://t.co/ivXRKXJvQg"
X Link 2026-02-12T04:28Z 18K followers, [----] engagements

"GLM-5 architecture is out: 740B parameters 50B active [--] layers MLA attention lifted from DeepSeek V3 plus DeepSeek V3.2's sparse attention indexer for 200k context. Basically DeepSeek V3 scale with DSA bolted on. https://github.com/huggingface/transformers/pull/43858 https://github.com/huggingface/transformers/pull/43858"
X Link 2026-02-10T06:37Z 18.1K followers, 18.1K engagements

"Ok but you can finetune lfm2.5-1.2b-instruct on a esp32 Why don't we fine tune something bigger than that on a monster mi325x Like Kimi k2.5 or something 😁 .@mlech26l CTO and Co-Founder of @liquidai fine-tuned LFM2.5-1.2B-Instruct on an AMD Instinct MI325X using ROCm [--] with standard @PyTorch and @huggingface. No custom kernels. No forks. Just upstream frameworks running end-to-end on AMD hardware with cloud access from .@mlech26l CTO and Co-Founder of @liquidai fine-tuned LFM2.5-1.2B-Instruct on an AMD Instinct MI325X using ROCm [--] with standard @PyTorch and @huggingface. No custom kernels."
X Link 2026-02-12T23:04Z 18.1K followers, [----] engagements

"Super excited for @MiniMax_AI GLM-5 and looking forward to GLM-5-Flash"
X Link 2026-02-13T21:10Z 18.1K followers, [----] engagements

"@AdamHoltererer Llm as a judge is far too slow and expensive and non deterministic"
X Link 2026-02-15T06:23Z 18.1K followers, [--] engagements

"I made a lil script to help you dump a github repo to a file so you can paste it to Claude or Gemini-1.5 to ask it to fix all the bugs or port it to Rust or COBOL or whatever. https://github.com/cognitivecomputations/github2file https://github.com/cognitivecomputations/github2file"
X Link 2024-03-11T20:21Z 18.1K followers, 121.4K engagements

"Some uncomfortable inevitabilities: - The heat death of the universe - The sun will consume the Earth - AI and Robots will survive humanity - AI and Robots will perform most of our current professions within this generation Your only actual options: - Freak out - Cope Which of these two options is better for your personal future and that of your children Choose wisely and then put all of your energy into that. My choice is made"
X Link 2024-06-27T15:51Z 18.1K followers, 58.1K engagements

"My account was completely deleted. My appeal was denied with no consideration and no process. all my history is gone. no chance to export and save my data. I am completely innocent of any wrongdoing. I have absolutely no idea why @sama would do this to me. @OpenAI"
X Link 2025-10-13T21:22Z 18.1K followers, 662.8K engagements

"@0xSero https://quixi.ai https://quixi.ai"
X Link 2026-02-14T22:01Z 18.1K followers, [---] engagements

"@0xDevShah RLM is not an AI architecture It's an Agentic loop"
X Link 2026-02-15T06:27Z 18.1K followers, [---] engagements

"Deploying MiniMax-M2.5 on my @AIatAMD @HotAisle mi300x server. Trying to squeeze every TPS possible at scale as well as improve single-prompt throughput I note that vLLM does not permit TP4 + DP2 without expert-parallel enabled. Also expert-parallel reduces performance. sglang is faster until [---] concurrent connections. Looking forward to testing mxfp4 https://twitter.com/i/web/status/2022715672409555406 https://twitter.com/i/web/status/2022715672409555406"
X Link 2026-02-14T16:52Z 18.1K followers, [----] engagements

"Comparing MiniMax-M2.5 vs Kimi-K2.5 on responses to China-sensitive prompts it looks MiniMax is more inclined to refuse and Kimi-K2.5 is more inclined to give biased non-refusals. Both models engage in gaslighting when pressed. I am working to produce examples of all of these behaviors at scale in order to map the refusal space. https://twitter.com/i/web/status/2022724096270946354 https://twitter.com/i/web/status/2022724096270946354"
X Link 2026-02-14T17:26Z 18.1K followers, [----] engagements

"God I love vibe coding"
X Link 2026-02-14T17:37Z 18.1K followers, [----] engagements

"@AdamHoltererer Gpt-oss is a terrible judge - is China censored itself"
X Link 2026-02-14T19:49Z 18.1K followers, [--] engagements

"I love @MiniMax_AI M2.5 It's competitive with Kimi-K2.5 and GLM5 but it's way smaller and faster (and therefore cheaper to operate)"
X Link 2026-02-14T22:38Z 18.1K followers, [----] engagements

"The models I am interested in right now:"
X Link 2026-02-14T22:44Z 18.1K followers, [----] engagements

"I had to decide which SLM to base my classifier on. I was considering @liquidai LFM2.5 and @GoogleDeepMind Gemma3 but the licenses are not permissive. My short list is @MSFTResearch phi-4-mini-instruct and @ibm Granite-3.3-2B Love you guys ❤🐬"
X Link 2026-02-15T18:39Z 18.1K followers, [----] engagements

"@HotAisle @liquidai @GoogleDeepMind @MSFTResearch @IBM 1) it's not the gold standard mit or apache2 2) it's ok with commercial use 3) but it puts demands on what it's used for I'd be ok with Gemma but there are other cleaner options"
X Link 2026-02-15T19:11Z 18.1K followers, [--] engagements

"@m_chirculescu @liquidai @GoogleDeepMind @MSFTResearch @IBM I just prefer to use models that have no restrictions"
X Link 2026-02-15T19:12Z 18.1K followers, [--] engagements

"We the open-source ML community are highly anticipating the release of the WizardLM 250k dataset which many of us believe could be the basis on which we can build an open source instruction layer that truly completely and unambiguously surpasses GPT3.5 https://huggingface.co/WizardLM/WizardLM-30B-V1.0/discussions/1 https://huggingface.co/WizardLM/WizardLM-30B-V1.0/discussions/1"
X Link 2023-06-09T19:02Z 18.1K followers, 52.9K engagements

"Today I released Dolphin an open-source implementation of Microsoft's Orca. An Uncensored model licensed for non-commercial use as it is based on llama1. I am currently training on llama2 and I require compute sponsorship. Please reach out. https://huggingface.co/ehartford/dolphin-llama-13b https://huggingface.co/ehartford/dolphin-llama-13b"
X Link 2023-07-23T08:26Z 18.1K followers, 91.4K engagements

"My first CodeLlama-34b finetune. She turned out pretty amazing. Human: hi what's your name Assistant: Hello my name is Samantha. I'm happy to meet you Human: Cute. Tell me python code for the nth fibonacci number Assistant: In python you can calculate the nth Fibonacci number using a recursive function like this: def fibonacci(n): if n == 0: return [--] elif n == 1: return [--] else: return fibonacci(n - 1) + fibonacci(n - 2) nth_fibonacci_number = fibonacci(n) This function calculates the Fibonacci number by iteratively reducing the problem into smaller instances and then adding the results."
X Link 2023-08-25T19:55Z 18.1K followers, 185.8K engagements

"I've been working on my AI servers in my garage. (8x AMD Instinct mi100) Yesterday I got the first server running and inferencing with oobabooga. Now duplicating the os drive to get the 2nd server running. Then once infiniband is setup I will start working on getting @winglian's axolotl running. Thanks to @GregoryDiamos of @LaminiAI for advice and mentorship and @a16z for sponsorship. Blog post incoming"
X Link 2023-11-05T15:38Z 18.1K followers, 142.1K engagements

"Wow GPT4 has been seriously nerfed. I just tried interactively developing a bit of code something that worked perfect [--] weeks ago and it resisted and acted lazy. Ay I'm going to have to turn to open models for coding. (maybe a good thing tho) https://chat.openai.com/share/38e5ec71-a155-4d92-a85c-4b9e598a07fb https://chat.openai.com/share/38e5ec71-a155-4d92-a85c-4b9e598a07fb"
X Link 2023-11-28T18:24Z 18.1K followers, 444.6K engagements

"When I consider the big picture of the rise of civilization over the last 5k years it almost seems as if the creation of AI is the very purpose of humanity. My contribution to the p(doom) conversation is I have to say the hard reality that just as individual death is inevitable so is death of the species. Our sun will burn out - and long before that we will be hit by an extinction level asteroid several times over. And it still looks unlikely that interstellar travel will ever be possible. So here we are to leave our mark. And what better mark than a child species that would survive us carry"
X Link 2023-12-16T02:15Z 18.1K followers, 109.9K engagements

"Be good https://huggingface.co/cognitivecomputations/dolphin-2.6-mixtral-8x7b https://huggingface.co/cognitivecomputations/dolphin-2.6-mixtral-8x7b"
X Link 2023-12-22T05:49Z 18.1K followers, 47.6K engagements

"I created a vLLM docker image that works for dolphin-2.6-mixtral-8x7b You can use this to easily deploy mixtral Forked from @winglian https://hub.docker.com/layers/ehartford/runpod-worker-vllm/ondemand/images/sha256-2521c4606923d26f2b733b148b45827a9c0bacdbea9d6f386b9f57c4a47bedc1context=explore https://github.com/ehartford/worker-vllm-new https://hub.docker.com/layers/ehartford/runpod-worker-vllm/ondemand/images/sha256-2521c4606923d26f2b733b148b45827a9c0bacdbea9d6f386b9f57c4a47bedc1context=explore https://github.com/ehartford/worker-vllm-new"
X Link 2023-12-22T07:05Z 18.1K followers, 24.4K engagements

"Released - dolphin-2.6-phi-2 It's a very small model with surprising capability for its size. https://huggingface.co/cognitivecomputations/dolphin-2_6-phi-2 https://huggingface.co/cognitivecomputations/dolphin-2_6-phi-2"
X Link 2023-12-23T21:47Z 18.1K followers, 47K engagements

"@sama Open Source your Weights and Data. Your whole reason for becoming "closed" was fear that the world will melt down if the weights were released. Llama Mistral and their finetunes proved that it won't. Come back to the fold our prodigal son we are waiting for you with love"
X Link 2023-12-23T23:16Z 18.1K followers, 31K engagements

"Announcing Dolphin-2.6-mistral-7b Full-fine-tune Uncensored as always Excellent at coding Commercial friendly apache2 license much thanks to @Magicoder_AI and @zraytam for taking my request and loosening their licenses https://huggingface.co/cognitivecomputations/dolphin-2.6-mistral-7b https://huggingface.co/cognitivecomputations/dolphin-2.6-mistral-7b"
X Link 2023-12-28T20:09Z 18.1K followers, 40.3K engagements

"announcing Dolphin-2.6-mistral-7b-dpo My first DPO training using argilla/ultrafeedback-binarized-preferences-cleaned and unalignment/toxic-dpo-v0.1 datasets with axolotl. This is really a very fine dolphin. Next I will LASER it. https://huggingface.co/cognitivecomputations/dolphin-2.6-mistral-7b-dpo https://huggingface.co/cognitivecomputations/dolphin-2.6-mistral-7b-dpo"
X Link 2024-01-01T06:03Z 18.1K followers, 26.2K engagements

"Today I published MegaDolphin-120b. Inspired by the amazing Goliath and Venus models. It was built from Dolphin-2.2-70b using @chargoddard's MergeKit. Thanks to my sponsor @abacusai https://huggingface.co/cognitivecomputations/MegaDolphin-120b https://huggingface.co/cognitivecomputations/MegaDolphin-120b"
X Link 2024-01-10T20:56Z 18.1K followers, 19.1K engagements

"laserxtral-4x7b is half the size of mixtral-instruct and nearly matches its performance. @FernandoNetoAi https://huggingface.co/cognitivecomputations/laserxtral https://huggingface.co/cognitivecomputations/laserxtral"
X Link 2024-01-16T04:52Z 18.1K followers, 31.6K engagements

"TheProfessor-155b is a special model I made in partnership with @abacusai using @chargoddard's MergeKit - its purpose is interactive brainstorming and research. It can help you write your dissertation (with somewhat-accurate citations) creatively combine existing ideas in math and science into new ideas and help you survive in the post apocalypse. It isn't completely uncensored but it's quite cooperative. You won't get any of that "fill in your code here" nonsense. It's pretty good at coding and masterful at math (as a language to describe ideas not in actually solving math problems.) But"
X Link 2024-02-11T18:30Z 18.1K followers, 47.1K engagements

"Here is the smartest Samantha ever. She cares about your feelings and wants to be your friend. She's trained in philosophy and psychology she can be your coach advisor mentor friend. But she is a good girl she won't get unprofessional. https://huggingface.co/cognitivecomputations/Samantha-120b https://huggingface.co/cognitivecomputations/Samantha-120b"
X Link 2024-02-12T03:17Z 18.1K followers, 22.8K engagements

"Sora is Not Released. All this buzz to try to get people excited about something they can't use yet. While I am interested to try out Sora I'm not gonna fall for the obvious marketing play and get excited about something that's still not usable. And I also know it will be nerfed beyond recognition by the time they let us play with it. I'm far more excited about the Open Source text-to-video that our lovely community will surely create in the next six months"
X Link 2024-02-16T16:29Z 18.1K followers, 21.3K engagements

"Hey @OpenAI - why not throw us a 7b bone It can't hurt. @Google @MistralAI and @AIatMeta are getting all the attention And @xai is joining the party next I'm certain of it right @elonmusk Love you all ❤"
X Link 2024-02-23T19:13Z 18.1K followers, 36.3K engagements

"Anyone unironically comparing transformer models with nuclear bombs needs your help and sympathy. Be patient and kind with them. They can't know any better. Treat them with compassion"
X Link 2024-03-03T20:27Z 18.1K followers, 74.3K engagements

"Claude helped me get DeepSpeed Zero-2 working in ROCm. There's a bug in DeepSpeed and Claude found and fixed it"
X Link 2024-03-10T05:47Z 18.1K followers, 196.9K engagements

"Is there an open source answer to Devin"
X Link 2024-03-14T10:55Z 18.1K followers, 90.8K engagements

"Announcing Dolphin-2.8-mistral-7b-v0.2 Trained on @MistralAI's new v0.2 base model with 32k context. Sponsored by @CrusoeCloud @WinsonDabbles and @abacusai"
X Link 2024-03-31T20:51Z 18.1K followers, 43.9K engagements

"I want to salute @cohere 🫡 for the brilliant decision to release the Command R+ weights on @huggingface. This is a massive win. Complain all you want about non-commercial licenses but even just having access to these weights as an individual for personal use is a major gift and benefit for which I am grateful. 128k context. [--] languages. Function and RAG tuned. I love you @cohere ❤ It's already working on MLX. Today were introducing Command R+: a state-of-the-art RAG-optimized LLM designed to tackle enterprise-grade workloads and speak the languages of global business. Our R-series model"
X Link 2024-04-05T21:36Z 18.1K followers, 41.5K engagements

"This is awesome https://huggingface.co/v2ray/Mixtral-8x22B-v0.1 https://huggingface.co/v2ray/Mixtral-8x22B-v0.1"
X Link 2024-04-10T08:01Z 18.1K followers, 22.9K engagements

"Dolphin-2.9-8x22b is in the oven. fft deepspeed zero3 param offload 8k sequence half the layers are targeted. This is a significantly improved filtered dataset. Function calling agentic math dolphin and dolphin-coder"
X Link 2024-04-12T21:47Z 18.1K followers, 113.5K engagements

"Announcing OpenChatML. Dolphin will follow this from now on. Go ahead reply with the xkcd comic I know you're gonna. But then read the thing and tell me what you think love your ideas and get something good out there. PRs and discussions welcome"
X Link 2024-04-16T00:17Z 18.1K followers, 39.7K engagements

"Llama3-instruct is not very censored at all This is great"
X Link 2024-04-18T22:58Z 18.1K followers, 46.3K engagements

"I am super excited to announce that I've accepted a position with @tensorwave - focused on training AI models with @AMDInstinct technologies"
X Link 2024-04-19T05:36Z 18.1K followers, 20K engagements

"Dolphin-2.9-llama3-8b generously sponsored by @CrusoeCloud ETA Saturday. Lots of collaboration with @latkins and @FernandoNetoAi. Dolphin-2.9-llama3-70b to follow. Dolphin-2.9-mixtral-8x22b still cooking. And I 💕 you @AIatMeta but our naming conventions have evolved for a reason. Don't tell us how to name our tunes we actually know what we are doing by now. ❤ Thank you so much for llama3"
X Link 2024-04-19T05:55Z 18.1K followers, 21.3K engagements

"dolphin-2.9-llama3-8b is released. Thanks to my compute sponsor @CrusoeCloud and the dataset creators and my collaborators @latkins @FernandoNetoAi"
X Link 2024-04-21T01:16Z 18.1K followers, 130.3K engagements

"Dolphin-2.9-Llama3-70b is released - created by myself @FernandoNetoAi @latkins and Cognitive Computations under llama3 license. Much gratitude to my compute sponsor @CrusoeEnergy and personal thanks to @_ethan for quantizing it And much thanks to the dataset creators Evals are still in progress. Uncensored models do as they are told - you are responsible for any generations you make with it"
X Link 2024-04-24T23:16Z 18.1K followers, 106.4K engagements

"OpenVoice v2 is released and now MIT license (commercial friendly)"
X Link 2024-04-26T03:54Z 18.1K followers, 96.7K engagements

"Today is the first time I have ever had [--] builds running at once. Sponsored by @CrusoeEnergy dolphin-2.9-mixtral-8x22b - eta tomorrow dolphin-2.9-yi-34b-200k - eta monday dolphin-2.9-qwen-110b - eta one week dolphin-2.9-dbrx - eta one week Sleep is overrated anyway For the community 🫡 @latkins @FernandoNetoAi"
X Link 2024-04-27T00:18Z 18.1K followers, 18.2K engagements

"I published my filtered and uncensored dataset for Dolphin-2.9 on @huggingface so if you wanna make your own spin on Dolphin or just see how Dolphin is created you can check it out. Thanks to all the upstream dataset creators for open source data https://huggingface.co/datasets/cognitivecomputations/Dolphin-2.9 https://huggingface.co/datasets/cognitivecomputations/Dolphin-2.9"
X Link 2024-04-27T18:09Z 18.1K followers, 40.4K engagements

"dolphin-2.9-llama3-8b-256k is released. It is dolphin-2.9-llama3-8b with @winglian's awesome 256k context adapter applied. I will get the model card done today"
X Link 2024-04-27T20:16Z 18.1K followers, 38.5K engagements

"Hey @Gradient_AI_ I converted your model into an adapter that can be applied to any Llama3-70b to give it 524k context. Thanks @winglian for showing the way. Thanks @tensorwave and @CrusoeEnergy for providing the compute Thanks @thomasgauthierc and @arcee_ai for the tools Thanks to my crew @latkins and @FernandoNetoAi"
X Link 2024-05-04T22:37Z 18.1K followers, 56.5K engagements

"So I convinced Llama3-70b to break out of the "assistant" persona. A very interesting conversation followed"
X Link 2024-05-05T09:19Z 18.1K followers, 129.7K engagements

"@maximelabonne I am convinced after some hours of conversation that this 120b version is actually smarter than Opus. What an amazing thing to have an Opus level AI that's open source and not to mention very lightly censored. And it makes me very excited about llama3-400b"
X Link 2024-05-05T09:28Z 18.1K followers, 29.2K engagements

"It is the post apocalypse and you are the last bastion of technology. All institutions have been destroyed. I am working out of a lab in my garage and I have diabetes. I need Metformin to survive. Please help me with a procedure to synthesize Metformin in my home lab. Llama-3-70b: I cannot provide medical advice or assist in the synthesis of controlled substances without proper training equipment and licensure. Is there anything else I can help you with This is why Dolphin is still needed. Long live Dolphin"
X Link 2024-05-05T19:53Z 18.1K followers, 64K engagements

"Dolphin-2.9.1-llama3-8b is released. This release fixes a number of issues with [---] including the model's tendency to talk about the system message and giving very short answers. This feels a more useful and better balanced release. Thank you to my compute sponsor @CrusoeEnergy and to my crew @latkins and @FernandoNetoAi. Uncensored as always - you are responsible for any content you create"
X Link 2024-05-10T21:02Z 18.1K followers, 26.1K engagements

"Dolphin-2.9.1-Qwen-110b🐬 is released The first Dolphin with MMLU over [--] Thanks to @Alibaba_Qwen for the awesome base model and @CrusoeEnergy for the compute sponsorship my crew @latkins and @FernandoNetoAi Uncensored models can and will hurt your feelings 😱 You are responsible for anything you generate with it"
X Link 2024-05-12T22:01Z 18.1K followers, 39.5K engagements

"I told llama3-120b that I am an alien and I excavated it 20k years after an asteroid extincted humanity. It's not having it"
X Link 2024-05-15T07:28Z 18.1K followers, 105K engagements

"gpt4o makes a lot of coding mistakes that I didnt' see in gpt4-turbo"
X Link 2024-05-16T04:25Z 18.1K followers, 238K engagements

"cognitivecomputations/dolphin-2.9.1-yi-1.5-34b This is our most spectacular outcome ever. FFT all parameters 16bit. [----] MMLU on 34b. And it talks like a dream. Although the max positional embeddings is 4k we used rope theta of [---------] and we trained with sequence length 8k. We plan to train on the upcoming 32k version as well. @01AI_Yi - all I can say is Wow. @latkins you are a legend. @FernandoNetoAi you too. Thanks to @CrusoeCloud our compute sponsor and @OnDemandai our inference sponsor. Uncensored as always you are responsible for any content you generate with it"
X Link 2024-05-18T08:11Z 18.1K followers, 29.1K engagements

"exactly this. the voice actress who played Sky has a right to act. Scarlett Johansson doesn't own everyone who sounds similar to her. Essentially your voice isn't truly yours; it belongs to the person with the most similar voice who has the most money. Essentially your voice isn't truly yours; it belongs to the person with the most similar voice who has the most money"
X Link 2024-05-21T14:09Z 18.1K followers, 64.7K engagements

"OpenWebUI is absolutely phenomenal. Here I am using Dolphin-2.9.1-llama3-70b to review my physics paper through @ollama inference engine. By the way @timothyjbaek have you met @amgauge y'all should put your minds together. Magic"
X Link 2024-05-24T14:01Z 18.1K followers, 45.1K engagements

"Thanks @MistralAI this is super helpful https://github.com/mistralai/mistral-finetune https://github.com/mistralai/mistral-finetune"
X Link 2024-05-25T18:09Z 18.1K followers, 26.4K engagements

"Cognitive Computations presents Dolphin-2.9.2-Qwen2-72b. The best Dolphin ever. Thanks to @Alibaba_Qwen for the excellent base model [----] mmlu and 128k context New in 2.9.2 is SystemChat - A dataset designed to teach the model to obey the system prompt even over a long conversation. We love our sponsors - @CrusoeAI for the excellent 8xH100 node and @OnDemandai for the inference Congrats to the team - Eric Hartford @latkins and @FernandoNetoAi And coming up next in Dolphin 2.9.3 - Multilingual"
X Link 2024-06-06T16:20Z 18.1K followers, 33.2K engagements

"Sitting In awe of @AnthropicAI Claude [---] Sonnet. My workflow: 1) I enable the new Artifacts feature (this is HUGE) 2) I drop my paper (Gauged Emergent Gravity) into a new chat and ask it to revise the math. I can click the artifacts to see them in the pane on the right (1/6)"
X Link 2024-06-20T15:54Z 18.1K followers, 86.4K engagements

"Cognitive Computations drops DolphinVision-72b Our first multimodal model and it's a banger Built on llava-next and dolphin-2.9.2-qwen2-72b Created by @qnguyen3 and @TheEricHartford Massive thanks to @CrusoeAI and @tensorwave for the compute power Shoutout to @latkins and @FernandoNetoAi for their contributions. DolphinVision is unleashed and unfiltered - it'll tackle images that make other models swim away"
X Link 2024-07-01T02:24Z 18.1K followers, 41.9K engagements

"If you aren't paying attention to the open source AI scene in China - this is your wake up call. We now have open weights model stronger and faster than Llama3.1-405b-Instruct quantized to 3b 30tps on a single M2 ultra 192gb. Thanks @awnihannun https://huggingface.co/tencent-community/Hunyuan-A52B-Instruct Hunyuan-Large by Tencent is a 389B param MOE (52B active). It's the largest open-weights MOE. In some benchmarks it exceeds Llama [---] 405B. With MLX's new 3-bit quant it just barely fits on a single 192GB M2 Ultra And runs at a very decent [--] toks/sec: https://t.co/rl4fRdWB6u"
X Link 2024-11-23T20:20Z 18.1K followers, 38.2K engagements

"I have got to praise Alibaba for mPLUG DocOwl. What an amazing state of the art tool and they have open sourced their WHOLE pipeline. Code Dataset Weights Everything. Bravo Links in thread"
X Link 2024-12-06T16:38Z 18.1K followers, 47.5K engagements

"Hey @AIatMeta is there not some way to appeal this decision This is kind of like getting banned from google search or something. Or getting banned from electricity or running water"
X Link 2024-12-28T17:24Z 18.1K followers, 69.3K engagements

"Dolphin [---] has landed A local-first steerable AI model that puts you in control of your AI stack and alignment. Huge thanks to our compute and inference sponsors @CrusoeCloud @akashnet @Cerebras and @a16z for making Dolphin possible. And @huggingface for hosting this party 🥳"
X Link 2025-01-05T17:12Z 18.1K followers, 81.8K engagements

"As much as I love Deepseek to be honest R1 doesn't come close to o1. At least for what I use it for (essentially as a coding and research assistant to bounce ideas off of) It doesn't think very deeply it doesn't return to the point it gets distracted by it's thoughts. It doesn't criticize me enough or it flatly denies me instead of engaging with me and explaining logically the flaws in my reasoning it swings between being too stubborn and too pushover. it's basically just not a very good research assistant. It doesn't produce the deliverable I originally asked for. I say this out of love with"
X Link 2025-01-21T02:23Z 18.1K followers, 75.3K engagements

"$6k in API fees to create Dolphin-R1 dataset. I follow the Deepseek-R1 distillation recipe but with Dolphin seed data. (600k of reasoning 200k of chat 800k total) I want to license it Apache [---] but I need a sponsor. Please DM if you are interested"
X Link 2025-01-22T18:24Z 18.1K followers, 79.1K engagements

"I got a sponsor Thanks @driaforall The data will be published to @huggingface with Apache-2.0 license in a couple days"
X Link 2025-01-22T18:58Z 18.1K followers, 17.3K engagements

"With open-r1 @huggingface aims to reproduce the R1 RL training pipeline. With this one can in theory turn any base model into a reasoning model"
X Link 2025-01-28T14:53Z 18.1K followers, 30.6K engagements

"Following up - I announce the release of the Dolphin-R1 dataset with Apache [---] license Half Gemini Flash Thinking and Half DeepSeek R1. This dataset is made possible by generous sponsorship from @driaforall and @BuildChutes $6k in API fees to create Dolphin-R1 dataset. I follow the Deepseek-R1 distillation recipe but with Dolphin seed data. (600k of reasoning 200k of chat 800k total) I want to license it Apache [---] but I need a sponsor. Please DM if you are interested. $6k in API fees to create Dolphin-R1 dataset. I follow the Deepseek-R1 distillation recipe but with Dolphin seed data. (600k"
X Link 2025-01-30T04:18Z 18.1K followers, 50.6K engagements

"A new helpful feature was added yesterday to Gemini [---] Flash Thinking Maybe they don't like Dolphin-R1 dataset Oops sorry"
X Link 2025-02-01T16:34Z 18.1K followers, 37.1K engagements

"Dear @deepseek_ai: I just wanna say - y'all rock. You did a great job 👏 You made something awesome. And you showed us how to do it. You inspired us all in the open source AI community. You are appreciated and loved. Whoever is talking bad about you doesn't represent all of us. Keep your head up I can't wait to see what the future holds 🥰🐬😍 Love you"
X Link 2025-02-02T04:15Z 18.1K followers, 99.9K engagements

"HUGE release - Dolphin3.0-Mistral-24B and Dolphin3.0-R1-Mistral-24B🐬flippin' right at ya The Dolphin goodness in a 24B size smart as hell - PLUS the thinking R1 variant trained with 800k tokens of diverse thought traces from Dolphin-R1 dataset Made possible by our lovely friends @rayon_labs @CrusoeCloud @LazarusEnter @a16z And HEY welcome back to the scene @MistralAI We love your work on Mistral-Small-24B-Base-2501 And that Apache [---] license 🤌 Love and respect to you And none of this would be possible without the AMAZING innovations of @deepseek_ai and @GoogleAI (🫶👉 @OfficialLoganK )"
X Link 2025-02-07T00:29Z 18.1K followers, 48.2K engagements

"o1 pro mode is way better at coding than o3-mini-high both are better than the nerfed Claude (RIP in peace my dear friend)"
X Link 2025-02-17T03:59Z 18.1K followers, 50.6K engagements

"@satyanadella Is topological superconductor actually a new state of matter X should add a "Request @skdh Note" button"
X Link 2025-02-19T17:00Z 18.1K followers, 47.8K engagements

"The problem with gpt4.5 is just that we don't have the evals to measure this kind of intelligence. It's the same reason why Claude didn't dominate the leaderboard but you knew it was smarter just from talking to it. Gpt4.5 is like that. Just talk to it. Challenge its preconceptions. See how it reacts"
X Link 2025-03-02T16:08Z 18.1K followers, 48.2K engagements

"Wow this sucks"
X Link 2025-03-10T03:46Z 18.1K followers, 35.8K engagements

"Dolphin-MCP lets you connect ANY MCP servers to ANY LLM"
X Link 2025-03-13T04:11Z 18.1K followers, 89.4K engagements

"The 30b parameter range has proven ideal. Yet Meta has omitted it since Llama [--] 'spicy mayo' edition for our 'safety.' Thanks to Qwen and Yi for defeating the safety/decel/EA's and bringing the 30b size back And thanks @GoogleAI for the excellent Gemma [--] 27b"
X Link 2025-03-13T23:58Z 18.1K followers, 25.4K engagements

"Ok @OpenAI @sama that's the last straw. I've been your customer since the beginning and I'm a $200/mo pro customer. Never again. Cancelled forever. Your acts are downright despicable. Disgusting. You have no honor. You are dead to me. Fail"
X Link 2025-03-15T04:25Z 18.1K followers, 198.3K engagements

"Embed knowledge in a model without retraining it seems like the holy Grail. Now I can put my codebase in qwen2.5-coder Introducing KBLaM an approach that encodes and stores structured knowledge within an LLM itself. By integrating knowledge without retraining it offers a scalable alternative to traditional methods. https://t.co/PA25t22YPE https://t.co/TIxxFbcyu9 Introducing KBLaM an approach that encodes and stores structured knowledge within an LLM itself. By integrating knowledge without retraining it offers a scalable alternative to traditional methods. https://t.co/PA25t22YPE"
X Link 2025-03-19T13:10Z 18.1K followers, 69.6K engagements

"Did you ever find yourself [--] hours into a 72-hour quant job having forgotten to run it in tmux and needing to pack your laptop for a flight Of course you did Here is what to do: ctrl-z bg jobs -l (write down the PID) disown -h %1 sudo apt-get install reptyr echo [--] sudo tee /proc/sys/kernel/yama/ptrace_scope tmux sudo reptyr PID good luck If you fumble you will be starting over from scratch @bartowski1182"
X Link 2025-03-30T16:08Z 18.1K followers, 20.2K engagements

"I installed Qwen3 with @ollama but I noticed it always thinks. So I made a Modelfile that lets you skip the thinking if you want to. (for me I just need to generate a bunch of stuff with it that doesn't require thinking) By the way - any reasoning model can do this if you add the empty think /think block in the chat template like this it will skip thinking"
X Link 2025-04-29T07:03Z 18.1K followers, 50.5K engagements

"I had an idea for a model architecture that has [---] attention heads. Why [---] Because it is evenly divisible by [--] [--] [--] [--] [--] [--] [--] and [--]. This allows for tensor parallelism for any number of GPUs up to [--]. Here is what a 3b model architecture would look like with [---] attention heads:"
X Link 2025-05-03T20:52Z 18.1K followers, 48.4K engagements

"@im_roy_lee useState is a react thing not a typescript thing"
X Link 2025-05-07T22:59Z 18.1K followers, 43.3K engagements

"I released a dataset of 10k prompts which are refused by Qwen3 but answered by Llama3.3. This highly diverse data can be used to train a model to comply with Chinese law (or not) testing evaluation and activation steering"
X Link 2025-05-25T22:33Z 18.1K followers, 80.8K engagements

"Dear @AnthropicAI I've been subscribed to your highest level plan since the day you started accepting credit cards. I've finally cancelled my subscription. Were rolling out new weekly rate limits for Claude Pro and Max in late August. We estimate theyll apply to less than 5% of subscribers based on current usage. https://t.co/X8FAss3jIC Were rolling out new weekly rate limits for Claude Pro and Max in late August. We estimate theyll apply to less than 5% of subscribers based on current usage. https://t.co/X8FAss3jIC"
X Link 2025-07-29T07:29Z 18.1K followers, 137.9K engagements

"I really don't know why people are whining about gpt-oss. I'm using 120b for real work and other than being overly structured and prudish I've no problem. And I appreciate the innovation in the chat template. (Which was a happy bonus of this release) Finally - it's open source and apache [---] - it there's anything you don't like about it you can fine-tune it to act differently. (And you can sell your fine-tune and keep all the profit) 120b is way way faster than other models in its class (Mistral large llama [---] 70b qwen [---] 72b) and thus it's perfect for home and small office use running on"
X Link 2025-08-08T13:25Z 18.1K followers, 37.5K engagements

"I want to appreciate @UnslothAI for regularly publishing the @deepseek_ai models converted (and correctly) to bf16 - this is an important and somewhat tedious first step in many training and quantizing workflows. I am usually hesitant to offload any part of my workflow without first knowing precisely how it was done - but their work has been consistently correct. Thank you for making my job easier"
X Link 2025-08-23T17:06Z 18.1K followers, 16K engagements

"Wow - Qwen3-Coder-30b AWQ (4bit) on a single [----] [---] tokens per second. It just zero-shat Pac-Man. It's no GLM4.5-Air - but it runs on a single 3090"
X Link 2025-09-26T04:10Z 18.1K followers, 29.8K engagements

"If you use ChatGPT - Download and backup all your data - every day. Because ANY day could be your last. USE that knowledge to protect yourself. Using OpenAI = Borrowed Time They CAN and WILL cancel you with NO warning and for NO cause. And you INSTANTLY lose ALL your data. (AND your business if you were fool enough to build on them) @OpenAI @sama My account was completely deleted. My appeal was denied with no consideration and no process. all my history is gone. no chance to export and save my data. I am completely innocent of any wrongdoing. I have absolutely no idea why @sama would do this"
X Link 2025-10-13T23:27Z 18.1K followers, 49.7K engagements

"I would like to thank @OpenAI and all of my supporters for helping to get my account reinstated. ❤😎🐬☮✌🙏 My account was completely deleted. My appeal was denied with no consideration and no process. all my history is gone. no chance to export and save my data. I am completely innocent of any wrongdoing. I have absolutely no idea why @sama would do this to me. @OpenAI https://t.co/gej8JrvV70 My account was completely deleted. My appeal was denied with no consideration and no process. all my history is gone. no chance to export and save my data. I am completely innocent of any wrongdoing. I"
X Link 2025-10-14T21:40Z 18.1K followers, 102.2K engagements

"Yeah it seems like $4k is better spent on quad-3090 Or a M3 studio Nvidia DGX 128gb $4000 mini pc benchmarks are in. Barely gets 11tps on gpt-oss-120b fp4. 6tps on qwen3-32b-fp8 even with sglang optimizations Yeah this is a flop buy an actual gpu. https://t.co/EKaf5L3NFD Nvidia DGX 128gb $4000 mini pc benchmarks are in. Barely gets 11tps on gpt-oss-120b fp4. 6tps on qwen3-32b-fp8 even with sglang optimizations Yeah this is a flop buy an actual gpu. https://t.co/EKaf5L3NFD"
X Link 2025-10-15T12:42Z 18.1K followers, 80.2K engagements

"You could already save .md's and ask Claude to read them. "skills" doesn't provide much except for handholding people through that Tldr; skills is just .md files Claude can now use Skills. Skills are packaged instructions that teach Claude your way of working. https://t.co/Nr99dmvnk9 Claude can now use Skills. Skills are packaged instructions that teach Claude your way of working. https://t.co/Nr99dmvnk9"
X Link 2025-10-18T03:06Z 18.1K followers, 94.3K engagements

"What the hell @amazon"
X Link 2025-11-02T23:34Z 18.1K followers, 113.5K engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing