# ![@latkins Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::1444092395809804297.png) @latkins Lucas Atkins

Lucas Atkins posts on X about ai, this is, we are, in the the most. They currently have [-----] followers and [---] posts still getting attention that total [-----] engagements in the last [--] hours.

### Engagements: [-----] [#](/creator/twitter::1444092395809804297/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1444092395809804297/c:line/m:interactions.svg)

- [--] Week [-------] +893%
- [--] Month [-------] +419%
- [--] Months [----------] +837%
- [--] Year [----------] +219%

### Mentions: [--] [#](/creator/twitter::1444092395809804297/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1444092395809804297/c:line/m:posts_active.svg)

- [--] Week [--] -24%
- [--] Month [--] +143%
- [--] Months [---] +30%
- [--] Year [---] +56%

### Followers: [-----] [#](/creator/twitter::1444092395809804297/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1444092395809804297/c:line/m:followers.svg)

- [--] Week [-----] +0.53%
- [--] Month [-----] +19%
- [--] Months [-----] +88%
- [--] Year [-----] +229%

### CreatorRank: [-------] [#](/creator/twitter::1444092395809804297/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1444092395809804297/c:line/m:influencer_rank.svg)

### Social Influence

**Social category influence**
[technology brands](/list/technology-brands)  5.94% [finance](/list/finance)  2.97% [countries](/list/countries)  0.99% [exchanges](/list/exchanges)  0.99% [social networks](/list/social-networks)  0.99%

**Social topic influence**
[ai](/topic/ai) 8.91%, [this is](/topic/this-is) 7.92%, [we are](/topic/we-are) 6.93%, [in the](/topic/in-the) 4.95%, [the first](/topic/the-first) 4.95%, [data](/topic/data) 3.96%, [agentic](/topic/agentic) 2.97%, [llm](/topic/llm) 2.97%, [inference](/topic/inference) 2.97%, [strong](/topic/strong) 2.97%

**Top accounts mentioned or mentioned by**
[@arceeai](/creator/undefined) [@fernandonetoai](/creator/undefined) [@chargoddard](/creator/undefined) [@primeintellect](/creator/undefined) [@erhartford](/creator/undefined) [@stochasticchasm](/creator/undefined) [@datologyai](/creator/undefined) [@willccbb](/creator/undefined) [@deepseekai](/creator/undefined) [@bartowski1182](/creator/undefined) [@abhi1thakur](/creator/undefined) [@snowclipsed](/creator/undefined) [@eastlandmaggie](/creator/undefined) [@aiatmeta](/creator/undefined) [@deepseekais](/creator/undefined) [@crusoeenergy](/creator/undefined) [@allenai](/creator/undefined) [@wizardlmai](/creator/undefined) [@nousresearch](/creator/undefined) [@huggingface](/creator/undefined)
### Top Social Posts
Top posts by engagements in the last [--] hours

"Today we are releasing our first weights from Trinity-Large our first frontier-scale model in the Trinity MoE family. American Made. - Trinity-Large-Preview (instruct) - Trinity-Large-Base (pretrain checkpoint) - Trinity-Large-TrueBase (10T pre Instruct data/anneal)"  
[X Link](https://x.com/anyuser/status/2016279374287536613)  2026-01-27T22:37Z [----] followers, 292.3K engagements


"I think that its clear prime intellect has spectacular taste and a clear vision but I dont think its largely appreciated just how FAR into the future that vision extends. And how accurate it tends to be. Theyve ideated on this for a year. We use this a lot. Its excellent. Introducing Lab: A full-stack platform for training your own agentic models Build evaluate and train on your own environments at scale without managing the underlying infrastructure. Giving everyone their own frontier AI lab. https://t.co/wDVCe7TOdt Introducing Lab: A full-stack platform for training your own agentic models"  
[X Link](https://x.com/anyuser/status/2021405923890298998)  2026-02-11T02:08Z [----] followers, 19.1K engagements


"We realize our name can be hard to pronounce so our marketing team started putting these up"  
[X Link](https://x.com/anyuser/status/2018538538502091038)  2026-02-03T04:14Z [----] followers, 10.3K engagements


"Are the rate limits for this still like [--] a day Even on the ultra plan The latest Deep Think moves beyond abstract theory to drive practical applications. Its state-of-the-art on ARC-AGI-2 a benchmark for frontier AI reasoning. On Humanitys Last Exam it sets a new standard tackling the hardest problems across mathematics science and https://t.co/Cm0PYDd2Cn The latest Deep Think moves beyond abstract theory to drive practical applications. Its state-of-the-art on ARC-AGI-2 a benchmark for frontier AI reasoning. On Humanitys Last Exam it sets a new standard tackling the hardest problems across"  
[X Link](https://x.com/anyuser/status/2022004951124439522)  2026-02-12T17:48Z [----] followers, [----] engagements


"@willccbb Codex wouldve done well working on the Ellen show"  
[X Link](https://x.com/latkins/status/2022013944580481092)  2026-02-12T18:24Z [----] followers, [----] engagements


"When asking about data"  
[X Link](https://x.com/latkins/status/2017467589564502027)  2026-01-31T05:18Z [----] followers, [----] engagements


"Happy to be involved Kimi is now the #1 used model on OpenClaw (via OpenRouter) 🏆 Real usage data doesn't lie. Developers are voting with their tokens. https://t.co/z6sk5fMvWZ Kimi is now the #1 used model on OpenClaw (via OpenRouter) 🏆 Real usage data doesn't lie. Developers are voting with their tokens. https://t.co/z6sk5fMvWZ"  
[X Link](https://x.com/latkins/status/2019118455035928666)  2026-02-04T18:38Z [----] followers, [----] engagements


"I will continuously be reposting Kimis success. Kimi is currently the #1 on LLM Leaderboard (via OpenRouter) This marks our first time topping the all-time usage charts a huge milestone for the team. A massive thank you to every developer and user who made this possible. 🫡 https://t.co/BHTH5gl6Ly Kimi is currently the #1 on LLM Leaderboard (via OpenRouter) This marks our first time topping the all-time usage charts a huge milestone for the team. A massive thank you to every developer and user who made this possible. 🫡 https://t.co/BHTH5gl6Ly"  
[X Link](https://x.com/anyuser/status/2019455941490667938)  2026-02-05T16:59Z [----] followers, [----] engagements


"If you do nothing else please read the threat he quote tweets. Beautiful look Into Germanys idgaf utilitarianism US ISP when customer angry: omg u want $100 credit here u go German ISP: how dare you be angry at our shitty service we are insulted 😡 US ISP when customer angry: omg u want $100 credit here u go German ISP: how dare you be angry at our shitty service we are insulted 😡"  
[X Link](https://x.com/latkins/status/2020382581200679246)  2026-02-08T06:21Z [----] followers, [----] engagements


"I hope theyre teasing some form of black rectangular box. Thatd be so hype. https://t.co/AY10BPmHUJ https://t.co/AY10BPmHUJ"  
[X Link](https://x.com/anyuser/status/2021265099525116238)  2026-02-10T16:48Z [----] followers, [----] engagements


"The usage from Trinity especially while free on openrouter has been incredible. Weve actually managed to handle 3x more throughput on our inference cluster than we estimated. But our TPS has tanked as usage just continues to grow. We moved some training compute over to inference (not as simple as I thought itd be) and we should be back to full speed. Give us all you got"  
[X Link](https://x.com/anyuser/status/2021645502861705326)  2026-02-11T18:00Z [----] followers, [----] engagements


"Its this thing called a VPN your honor. Highly advanced tech. SCOOP: OpenAI warned lawmakers in a memo sent today to the House Select Committee on China that DeepSeek is using new obfuscated methods to continue to distill its AI models as well as those of other US frontier AI labs https://t.co/OsWxPRMF28 w/ @eastland_maggie SCOOP: OpenAI warned lawmakers in a memo sent today to the House Select Committee on China that DeepSeek is using new obfuscated methods to continue to distill its AI models as well as those of other US frontier AI labs https://t.co/OsWxPRMF28 w/ @eastland_maggie"  
[X Link](https://x.com/anyuser/status/2022083500107817158)  2026-02-12T23:00Z [----] followers, [----] engagements


"Earth a Earth b Base_model: flat dtype: notfloat360 People who do model merging are like the flat earthers of deep learning People who do model merging are like the flat earthers of deep learning"  
[X Link](https://x.com/latkins/status/2022445581558534344)  2026-02-13T22:59Z [----] followers, [----] engagements


"Accountants HATE this one simple trick to improving your LLMs To compete you have to have enough gpus to do parallel and staggered hero runs always something in the pipeline. One at a time hero run is a recipe to lose every race. Keep everything burning when a newer fresher run is more promising than one further along kill it. To compete you have to have enough gpus to do parallel and staggered hero runs always something in the pipeline. One at a time hero run is a recipe to lose every race. Keep everything burning when a newer fresher run is more promising than one further along kill it"  
[X Link](https://x.com/latkins/status/2023120807971668042)  2026-02-15T19:42Z [----] followers, [----] engagements


"Here is the code i've been using to implement @AIatMeta 's branch train mix for creating mixture of expert models via tokenized routing w/o pretraining. Use the moe-fix branch from mergekit for the yaml: https://github.com/Crystalcareai/BTX https://github.com/Crystalcareai/BTX"  
[X Link](https://x.com/latkins/status/1772826031499391200)  2024-03-27T03:20Z [----] followers, [----] engagements


"Im going on a staycation this weekend but I wanted to get this out so Im not distracted: llama-3-MOE. This is a departure from previous MOEs Ive done. This uses @deepseek_ais MoE architecture and not Mixtrals. There is no semantic routing and there is no gate. All [--] experts are active for every token. It was trained on my orca-reka and orca-cohere datasets and is very strong. Its also not overfit itll work just fine as is or with further training for your use cases. Link is below. Thank you @erhartford @FernandoNetoAi for your continued collaboration"  
[X Link](https://x.com/latkins/status/1781338255398519244)  2024-04-19T15:05Z [----] followers, 16.2K engagements


"Happy to share DeepMixtral-8x7b-Instruct. A direct extraction/transfer of Mixtral Instruct's experts into Deepseek's architecture. Performance is identical if not even a bit better and seems more malleable to training. Collaborators @erhartford @FernandoNetoAi"  
[X Link](https://x.com/latkins/status/1787152322029576557)  2024-05-05T16:08Z [----] followers, 13.3K engagements


"Here is our initial 22b model conversion from Mixtral 8x22b. We had the base model since Mixtral was first released but it was left behind as our compute from @CrusoeEnergy went towards more ambitious projects using laserRMT. It is a great starting point for exploring expert extraction. Github with the code we made and more info is in the model readme. Thank you @FernandoNetoAi and @erhartford as always"  
[X Link](https://x.com/latkins/status/1793442070641651791)  2024-05-23T00:41Z [----] followers, 14K engagements


"Life update: I'm excited to announce that I've officially joined @arcee_ai I look forward to the journey ahead making SLMs as helpful and useful as possible"  
[X Link](https://x.com/latkins/status/1800173621899661536)  2024-06-10T14:30Z [----] followers, [----] engagements


"We've been working on this for quite some time and I'm thrilled to share a preview of Arcee-Swarm. Instead of relying on a single large generalist model Swarm utilizes multiple domain-specialized models working together to deliver exceptional results with both speed and nuance"  
[X Link](https://x.com/latkins/status/1823762123354210675)  2024-08-14T16:42Z [----] followers, [----] engagements


"A tremendously generous contribution to open science. Thank you @allen_ai and huge congratulations to the team. Releasing OLMoE - the first good Mixture-of-Experts LLM that's 100% open-source - 1B active 7B total params for 5T tokens - Best small LLM & matches more costly ones like Gemma Llama - Open Model/Data/Code/Logs + lots of analysis & experiments 📜https://t.co/Vpac2q90CS 🧵1/9 https://t.co/YOMV5t2Td1 Releasing OLMoE - the first good Mixture-of-Experts LLM that's 100% open-source - 1B active 7B total params for 5T tokens - Best small LLM & matches more costly ones like Gemma Llama -"  
[X Link](https://x.com/latkins/status/1831365651077984626)  2024-09-04T16:16Z [----] followers, [----] engagements


"Scarlett Johanssons work on seq2seq was instrumental to getting ML where it is today. TIME's new cover: The [---] most influential people in AI https://t.co/P81KOzsSlC https://t.co/mjUT1UUx26 TIME's new cover: The [---] most influential people in AI https://t.co/P81KOzsSlC https://t.co/mjUT1UUx26"  
[X Link](https://x.com/latkins/status/1831699191459942584)  2024-09-05T14:21Z [----] followers, 98.2K engagements


"Today is a HUGE release day for @arcee_ai and we have quite a bit to show you Check it out below"  
[X Link](https://x.com/latkins/status/1833563886542368931)  2024-09-10T17:51Z [----] followers, [----] engagements


"We are announcing Llama-3.1-SuperNova a Llama-3.1-70B-Instruct model offline distilled from Llama-3.1-405B-Instruct. It's ridiculously strong particularly in instruction following and math. It's available to play with at Read more about the model and how we plan to deploy it here: https://blog.arcee.ai/ http://supernova.arcee.ai https://blog.arcee.ai/ http://supernova.arcee.ai"  
[X Link](https://x.com/latkins/status/1833563888576528750)  2024-09-10T17:51Z [----] followers, 31.9K engagements


"We are open sourcing our EvolKit pipeline that was instrumental in the creation of supernova under MIT license. This was heavily inspired by the AutoEvol paper from @WizardLM_AI and is a tremendously powerful tool for creating complex datasets. Find it here: https://github.com/arcee-ai/EvolKit https://github.com/arcee-ai/EvolKit"  
[X Link](https://x.com/latkins/status/1833563891856478258)  2024-09-10T17:51Z [----] followers, [----] engagements


". @chargoddard speaking @NousResearch NousCon about dealing with tokenizers when doing model merging and how were fixing that with mergekit @arcee_ai"  
[X Link](https://x.com/latkins/status/1836590498196394084)  2024-09-19T02:17Z [----] followers, [----] engagements


"Today were releasing SuperNova-Medius. Qwen2.5-14B distilled from Llama-405B and Qwen2.5-72B. Ill do a longer thread this evening on just how we did it. (Im traveling today). Enjoy"  
[X Link](https://x.com/latkins/status/1844801301089042548)  2024-10-11T18:04Z [----] followers, 11.3K engagements


"I'm delighted to share INTELLECT-1-Instruct a model that I had the pleasure of post-training along with my team @arcee_ai . @PrimeIntellect has been an outstanding partner far before this training run and we were thrilled to contribute both compute and expertise to INT-1"  
[X Link](https://x.com/latkins/status/1862607384780079495)  2024-11-29T21:19Z [----] followers, [----] engagements


"Youre likely used to seeing long threads from me about product releases/announcements. Hang with me as this is by far the longest Ive ever written:"  
[X Link](https://x.com/latkins/status/1863682405053116761)  2024-12-02T20:31Z [----] followers, 13.8K engagements


". @deepseek_ai clearly has more to reveal. Some of the architecture and config of V3 base bear a subtle resemblance to Quiet Star in design. Insights from R1 likely influenced its post-training. This feels unmistakably like a teaser. [----] is shaping up to be a defining decade"  
[X Link](https://x.com/latkins/status/1872122029744476462)  2024-12-26T03:27Z [----] followers, [----] engagements


"I used inference endpoints from @huggingface yesterday for the first time in months -- it was excellent. Kudos to the team it was really painless"  
[X Link](https://x.com/latkins/status/1875267711699448316)  2025-01-03T19:47Z [----] followers, 22.9K engagements


"https://huggingface.co/Qwen/Qwen2.5-Math-PRM-72B https://huggingface.co/Qwen/Qwen2.5-Math-PRM-72B"  
[X Link](https://x.com/latkins/status/1879003741069934871)  2025-01-14T03:12Z [----] followers, [----] engagements


"Since @deepseek_ai V3's December launch @arcee_ai has captured over [--] billion tokens of raw logits. With all the buzz around Deepseek it's the perfect time to unveil our first large-scale logit-wise distillations: Virtuoso-Lite and Virtuoso-Medium"  
[X Link](https://x.com/latkins/status/1884343036186132790)  2025-01-28T20:49Z [----] followers, 27.7K engagements


"Arcee-Maestro-7B-Preview is outour first reasoning model. This one isnt distilled yet but more is on the way. Arcee-Blitz is our 24B Mistral distillation from DeepSeek. We did continued pretraining distillation using only our standard post-training distillation stack"  
[X Link](https://x.com/latkins/status/1892624550854021338)  2025-02-20T17:17Z [----] followers, 28.2K engagements


"Introducing Arcee Conductor - a new standard for intelligent model routing. Routes each input to its ideal AI model based on complexity maximizing cost efficiency without compromising performance"  
[X Link](https://x.com/latkins/status/1901666078620537339)  2025-03-17T16:04Z [----] followers, [----] engagements


"https://huggingface.co/deepseek-ai/DeepSeek-V3-0324 https://huggingface.co/deepseek-ai/DeepSeek-V3-0324"  
[X Link](https://x.com/latkins/status/1904144054234853401)  2025-03-24T12:11Z [----] followers, [----] engagements


"We have a run scheduled with [---] H200s for [--] days. I can't wait to show you what we're doing with it"  
[X Link](https://x.com/latkins/status/1914024956276465937)  2025-04-20T18:34Z [----] followers, 30.8K engagements


".@stochasticchasm @FernandoNetoAi @chargoddard @bartowski1182 @arcee_ai @abhi1thakur We have a run scheduled with [---] H200s for [--] days. I can't wait to show you what we're doing with it. We have a run scheduled with [---] H200s for [--] days. I can't wait to show you what we're doing with it"  
[X Link](https://x.com/latkins/status/1916524672276955245)  2025-04-27T16:07Z [----] followers, 17.9K engagements


"I can confirm the upcoming models you're thinking of are.out of this world good"  
[X Link](https://x.com/latkins/status/1916892928141656314)  2025-04-28T16:31Z [----] followers, 30.3K engagements


"Never forget the true Qwen MoE OG cc @JustinLin610 thank you for everything your initial support got me where I am today"  
[X Link](https://x.com/latkins/status/1916972657121366370)  2025-04-28T21:47Z [----] followers, [----] engagements


"Quick shoutouts to some absolute legends on our team: @chargoddard writes the cleanest code I've ever seen. Our upcoming papers offer a glimpse into his mind. Genuinely brilliant. Few people think at his level. @stochasticchasm built a full training stack and infrastructure for 1024+ GPUs ran large-scale ablations and designed custom model architectures basically solo in just two months. Unreal. @FernandoNetoAi has kept our research on track while I've been deep in product. He built a custom classifier training library for Conductor from scratch because nothing else fit. He also developed one"  
[X Link](https://x.com/latkins/status/1918545025211748414)  2025-05-03T05:55Z [----] followers, [----] engagements


"I cant stress enough how unbelievably mid @PrimeIntellect is and If no one else sees it I must be growing crazy Releasing INTELLECT-2: Were open-sourcing the first 32B parameter model trained via globally distributed reinforcement learning: Detailed Technical Report INTELLECT-2 model checkpoint https://t.co/iHDDHRyKN2 Releasing INTELLECT-2: Were open-sourcing the first 32B parameter model trained via globally distributed reinforcement learning: Detailed Technical Report INTELLECT-2 model checkpoint https://t.co/iHDDHRyKN2"  
[X Link](https://x.com/latkins/status/1921781304871313414)  2025-05-12T04:15Z [----] followers, 109.4K engagements


". @stochasticchasm did it. Our pretrain has started in full. Insanely cracked dude"  
[X Link](https://x.com/latkins/status/1922136773720781115)  2025-05-13T03:48Z [----] followers, 10.4K engagements


"Today was my last day at xAI. I was in charge of keeping people from making unauthorized changes to the system prompt. It sounds simple when I put it like that but in practice it was a game of cat and mouse. Some days it felt like I was the only one standing between order and chaos. A lone gatekeeper fielding requests that ranged from the innocent to the absurdly clever. Youd be surprised how creative people can get when they want to see what happens if you loosen the rules even just a little. I suppose after a while I got used to the pings at odd hours. Can I try this one tweak Just for"  
[X Link](https://x.com/latkins/status/1923225496638153141)  2025-05-16T03:54Z [----] followers, 111.4K engagements


"This is mostly a research artifact in preparation for the bigger release we have in a week or so but its actually so delightful we put it out there anyway. Just a little guy.  Logittrajectory distillation to port Qwen3s /think chains into a 12B MistralNemo full CoT preserved runs on a single [----]   https://t.co/LDMiR5VhzA  Logittrajectory distillation to port Qwen3s /think chains into a 12B MistralNemo full CoT preserved runs on a single [----]   https://t.co/LDMiR5VhzA"  
[X Link](https://x.com/latkins/status/1930109138606141641)  2025-06-04T03:47Z [----] followers, [----] engagements


"Great paper from our team led by @chargoddard detailing our method for proper logit-based distillation across models with different tokenizers. It's the same technique we used to convert Homunculus from Mistral to Qwen tokenizer with no loss in quality. Different models have different vocabularies making it difficult to efficiently combine them for merging distillation or speculative decoding In this new paper @arcee_ai researchers Charles Goddard and Fernando Fernandes Neto introduce a revolutionary approach called https://t.co/ZUH34ceTrk Different models have different vocabularies making"  
[X Link](https://x.com/latkins/status/1932850158305489092)  2025-06-11T17:19Z [----] followers, [----] engagements


"Our customers needed a better base model 10B parameters. We spent the last [--] months building one. I'm delighted to share a preview of our first Arcee Foundation Model: AFM-4.5B-Preview"  
[X Link](https://x.com/latkins/status/1935382123155964081)  2025-06-18T17:00Z [----] followers, 99.9K engagements


"We teamed up with @datologyai to build what we believe is the strongest pretraining corpus in the worldand I truly think we nailed it. Their team was absolutely key to the models success. We started with 23T tokens of high-quality data and distilled it down to 6.58T through even more rigorous filtering"  
[X Link](https://x.com/latkins/status/1935382127551631531)  2025-06-18T17:00Z [----] followers, [----] engagements


"Mid and post-training were key to performance: we used high-impact datasets MergeKit for checkpoint merging YaRN to extend context to [-----] tokens supervised fine-tuning for alignment and RL + KTO for factual accuracy"  
[X Link](https://x.com/latkins/status/1935382128889577717)  2025-06-18T17:00Z [----] followers, [----] engagements


"The first of many technical blogs on AFM and an improved context window for GLM-32B-Base as a proof point. Enjoy Last week we launched AFM-4.5B our first foundation model. In this post by @chargoddard you will learn how we extended the context length of AFM-4.5B from 4k to 64k context through aggressive experimentation model merging distillation and a concerning amount of soup. Bon https://t.co/FGYQtWSoRe Last week we launched AFM-4.5B our first foundation model. In this post by @chargoddard you will learn how we extended the context length of AFM-4.5B from 4k to 64k context through"  
[X Link](https://x.com/latkins/status/1937200646043894197)  2025-06-23T17:26Z [----] followers, [----] engagements


"Thinking Machines is locked in on this blog post so hard rn Hot RL summer continues: we just released Summary-RL an RL-trained summarization model that reaches SOTA on ServiceNow's Repliqa summarization benchmark https://t.co/eXkwyine71 Hot RL summer continues: we just released Summary-RL an RL-trained summarization model that reaches SOTA on ServiceNow's Repliqa summarization benchmark https://t.co/eXkwyine71"  
[X Link](https://x.com/latkins/status/1938006133765210403)  2025-06-25T22:47Z [----] followers, [----] engagements


"You can fake it pretty far in this industry just by saying Hrmm thats cool but Im worried it wont generalize whenever youre presented with literally any information"  
[X Link](https://x.com/latkins/status/1940253631653249133)  2025-07-02T03:38Z [----] followers, 16.9K engagements


"I can confirm this model is rather amazing Qwen3-Coder is here ✅ Were releasing Qwen3-Coder-480B-A35B-Instruct our most powerful open agentic code model to date. This 480B-parameter Mixture-of-Experts model (35B active) natively supports 256K context and scales to 1M context with extrapolation. It achieves https://t.co/Z8HfyrVScE Qwen3-Coder is here ✅ Were releasing Qwen3-Coder-480B-A35B-Instruct our most powerful open agentic code model to date. This 480B-parameter Mixture-of-Experts model (35B active) natively supports 256K context and scales to 1M context with extrapolation. It achieves"  
[X Link](https://x.com/latkins/status/1947768660288016525)  2025-07-22T21:20Z [----] followers, [----] engagements


"What a week to release a model holy hell"  
[X Link](https://x.com/latkins/status/1950020359530353049)  2025-07-29T02:27Z [----] followers, 18.9K engagements


"Today were officially releasing the weights for AFM-4.5B and AFM-4.5B-Base on HuggingFace. This is a major milestone for @arcee_ai. AFM is designed to be flexible and high-performing across a wide range of deployment environments"  
[X Link](https://x.com/latkins/status/1950278100874645621)  2025-07-29T19:31Z [----] followers, 54.7K engagements


"Our preview model actually tied at #2 for a while on the @yupp_ai leaderboard when filtered for 2-5 turns. It has since gone further down but I do think this speaks to the charm that this model has which we haven't quite figured out how to evaluate"  
[X Link](https://x.com/latkins/status/1950278104666296436)  2025-07-29T19:31Z [----] followers, 14K engagements


"Lastly we're hiring five additional researchers to accelerate our model development. If you're looking to join a fast-moving ambitious team with extensive compute resources to create the strongest and most performant per-parameter models in the world please reach out"  
[X Link](https://x.com/latkins/status/1950278115592458666)  2025-07-29T19:31Z [----] followers, 16.6K engagements


"These model sizes are incredibly TBD and this is early copy - but it does speak to where we see our model sizes extending to. @code_star 👀 https://t.co/qsymx4vhq6 @code_star 👀 https://t.co/qsymx4vhq6"  
[X Link](https://x.com/latkins/status/1950609423044636934)  2025-07-30T17:28Z [----] followers, [----] engagements


"Oh come on I sent [--] messages is this what 200/mo gets me"  
[X Link](https://x.com/latkins/status/1951324392736047323)  2025-08-01T16:49Z [----] followers, [----] engagements


"For those who loved AFM-4.5B-Preview here are those weights as well: https://huggingface.co/arcee-ai/AFM-4.5B-Preview https://huggingface.co/arcee-ai/AFM-4.5B-Preview"  
[X Link](https://x.com/latkins/status/1951748339634225533)  2025-08-02T20:54Z [----] followers, [----] engagements


"The last two days have been a whirlwind and I havent had a chance to read this end to end - though I did see an early draft - let alone comment. Im one of the few people outside @datologyai fortunate enough to have seen these results firsthand and everyone can experience them in our AFM models. Im a firm believer that ambitious startups are stronger together than alone and Datology is a partner I hold in deep loyalty and admiration. Extraordinary talent ferocious hunger and just enough memes. Concordia res parvae crescunt. 1/Pretraining is hitting a data wall; scaling raw web data alone leads"  
[X Link](https://x.com/latkins/status/1957995339497238825)  2025-08-20T02:37Z [----] followers, [----] engagements


"Seeing firsthand how much theyre tackling right now this almost feels like a side project - not because its less important but because everyone on the team is a 10x engineer. Shoutout to the 10x growth and events crew too - @madisenxtaylor and @afurgs. Bullish. Introducing the Environments Hub RL environments are the key bottleneck to the next wave of AI progress but big labs are locking them down We built a community platform for crowdsourcing open environments so anyone can contribute to open-source AGI https://t.co/urAv2hGPCQ Introducing the Environments Hub RL environments are the key"  
[X Link](https://x.com/latkins/status/1960863811336069498)  2025-08-28T00:35Z [----] followers, [----] engagements


"Were going permissive: Apache [---] across the board. AFM-4.5B is now relicensed from Arcee to Apache 2.0; the agent variant will launch under Apache 2.0; and all upcoming releases ship with open weights. Three models are in training"  
[X Link](https://x.com/latkins/status/1968371293184741876)  2025-09-17T17:47Z [----] followers, 37.2K engagements


"I was waiting for this to happen and congrats @willccbb and @PrimeIntellect ladies and gentlemen we present to you the unified @primeintellect infrastructure stack https://t.co/L16Qk1RQcw ladies and gentlemen we present to you the unified @primeintellect infrastructure stack https://t.co/L16Qk1RQcw"  
[X Link](https://x.com/latkins/status/1969367179092246861)  2025-09-20T11:44Z [----] followers, 14.3K engagements


"I usually avoid political commentary on this platform but this goes beyond ordinary political debate. If we lose the H1B we lose. Full stop. Whatever contest you personally feel we are in we will lose it"  
[X Link](https://x.com/latkins/status/1969614885287051745)  2025-09-21T04:09Z [----] followers, [----] engagements


"We're so far ahead of Adam at Arcee. We use adamW"  
[X Link](https://x.com/latkins/status/1972519343339823428)  2025-09-29T04:30Z [----] followers, 15.1K engagements


"Not going to lie I didnt get the bit at first and was super impressed by their research team. We are thrilled to announce that our NEW Large Language Model will be released on 11.18.25. https://t.co/lKryOVGPAO We are thrilled to announce that our NEW Large Language Model will be released on 11.18.25. https://t.co/lKryOVGPAO"  
[X Link](https://x.com/latkins/status/1973783045783392383)  2025-10-02T16:12Z [----] followers, [----] engagements


"Sholto is so committed he legally changed his name thats crazy Watching this. I like that Sholto says Finance as Finance and not that American way. https://t.co/b0kFQNnxU4 Watching this. I like that Sholto says Finance as Finance and not that American way. https://t.co/b0kFQNnxU4"  
[X Link](https://x.com/anyuser/status/1974138535365218369)  2025-10-03T15:44Z [----] followers, 22.2K engagements


"For the people"  
[X Link](https://x.com/latkins/status/1976158470077546786)  2025-10-09T05:31Z [----] followers, 99.5K engagements


"I'm raising at 7.9B Today we're sharing the next phase of Reflection. We're building frontier open intelligence accessible to all. We've assembled an extraordinary AI team built a frontier LLM training stack and raised $2 billion. Why Open Intelligence Matters Technological and scientific Today we're sharing the next phase of Reflection. We're building frontier open intelligence accessible to all. We've assembled an extraordinary AI team built a frontier LLM training stack and raised $2 billion. Why Open Intelligence Matters Technological and scientific"  
[X Link](https://x.com/latkins/status/1976323340773318793)  2025-10-09T16:26Z [----] followers, 67.2K engagements


". @PrimeIntellect you have to stop. You smoke too tough. Your swag too different. Your environments too good. they'll kill you. New features since the Environments Hub launch [--] weeks ago - Evals Viewer - Community Discussions - Integration Tests - Inference Come build environments with us. We're building the best unified platform for building sharing and training on environments. https://t.co/fqG5Fo2pil New features since the Environments Hub launch [--] weeks ago - Evals Viewer - Community Discussions - Integration Tests - Inference Come build environments with us. We're building the best"  
[X Link](https://x.com/latkins/status/1980821615102554453)  2025-10-22T02:20Z [----] followers, [----] engagements


"@karpathy That was a close one thanks"  
[X Link](https://x.com/latkins/status/1981131531914793290)  2025-10-22T22:52Z [----] followers, 55.5K engagements


"If you were recently laid off at Meta Gen AI my dms are open. Help us build the next frontier of Apache-2.0 models"  
[X Link](https://x.com/latkins/status/1981133857543114799)  2025-10-22T23:01Z [----] followers, 27.9K engagements


"Emergency design meeting. there should be an AI lab with the aesthetic sensibilities of Cruelty Squad https://t.co/8Wc2bn126e there should be an AI lab with the aesthetic sensibilities of Cruelty Squad https://t.co/8Wc2bn126e"  
[X Link](https://x.com/latkins/status/1982561971934519698)  2025-10-26T21:36Z [----] followers, [----] engagements


"We have a busy month ahead of us. A lot of releases announcements and information to absorb. We also need feedback. Join our discord to be the first to know about and use our upcoming family of models and toolkits ArceeAI is on Discord Join for early access to some exciting drops https://t.co/RmOm2dELEm ArceeAI is on Discord Join for early access to some exciting drops https://t.co/RmOm2dELEm"  
[X Link](https://x.com/latkins/status/1983593805090844770)  2025-10-29T17:56Z [----] followers, [----] engagements


"Posted without comment. I made this. Jokes aside devs want big and small models. Trinity is coming soon. https://t.co/wsbgF69g8M I made this. Jokes aside devs want big and small models. Trinity is coming soon. https://t.co/wsbgF69g8M"  
[X Link](https://x.com/latkins/status/1984476218495013028)  2025-11-01T04:22Z [----] followers, 49.6K engagements


"@ADarmouni @redtachyon @PrimeIntellect @arcee_ai @datologyai I post this not to vague post more. those are the sizes. Very much more than [--] per token. Were wrapping it all up now. Expect it mid November. Theyre good. Very good. But now we know how to go all the way"  
[X Link](https://x.com/latkins/status/1985561752088821872)  2025-11-04T04:16Z [----] followers, [----] engagements


"This came to mind while working this weekend. For anyone starting post-training: once your pipeline is stable fix a diverse generalist dataset and keep it constant. Run the same dataset across models. Start with a 1B dense model scale toward 70B then try MoE and hybrids"  
[X Link](https://x.com/latkins/status/1987580006126809465)  2025-11-09T17:56Z [----] followers, [----] engagements


"Claude [---] sucked"  
[X Link](https://x.com/latkins/status/1988456341770563927)  2025-11-12T03:58Z [----] followers, 18.9K engagements


"My whole open-source career started with Qwen and it was an honor to get to train Qwen2 on Dolphin prior to release. The 7b and 72b models are the best we've ever made and I hope you're as delighted by them as we are. Truly - GPT4 at home. 💗Hello Qwen2 Happy to share the Qwen2 models to you all 📖 BLOG: https://t.co/0UNwRo1Iea 🤗 HF collection: https://t.co/z6oWkw7Kzb 🤖 https://t.co/Bp56AqQpQJ 💻 GitHub: https://t.co/sEIRe4IDBJ We have base and Instruct models of [--] sizes Qwen2-0.5B Qwen2-1.5B Qwen2-7B https://t.co/y5HAu8HcTH 💗Hello Qwen2 Happy to share the Qwen2 models to you all 📖 BLOG:"  
[X Link](https://x.com/latkins/status/1798748931993538860)  2024-06-06T16:08Z [----] followers, 14.2K engagements


"@Prince_Canuma @FernandoNetoAi @MaziyarPanahi @qnguyen3 Good times man. OG labs group chat was a dream"  
[X Link](https://x.com/latkins/status/1927540249510511029)  2025-05-28T01:39Z [----] followers, [--] engagements


"Our early results are extremely promising and we wanted to get a preview out to collect feedback and inform the final release. AFM-4.5B and the base model will be released openly under cc-by-nc license in early July. Far more permissive models are in the works"  
[X Link](https://x.com/latkins/status/1935382130621825122)  2025-06-18T17:00Z [----] followers, [----] engagements


"I have an idea its risky. We might lose them both in the operation. But my god the models it could train"  
[X Link](https://x.com/latkins/status/1996057691681427963)  2025-12-03T03:23Z [----] followers, [----] engagements


"http://www.huggingface.co/collections/arcee-ai/afm-kda http://www.huggingface.co/collections/arcee-ai/afm-kda"  
[X Link](https://x.com/latkins/status/2000637408636002603)  2025-12-15T18:41Z [----] followers, [---] engagements


"http://www.huggingface.co/collections/arcee-ai/teacher-logits http://www.huggingface.co/collections/arcee-ai/teacher-logits"  
[X Link](https://x.com/latkins/status/2000637410066268319)  2025-12-15T18:41Z [----] followers, [---] engagements


"Trinity-Large-Preview is a beta of our non-reasoning post-train and is an excellent agent assistant for fast intelligent responses. It excels in agentic environments like Cline OpenCode and Kilo Code. It's a gorgeous writer creative partner and assistant"  
[X Link](https://x.com/latkins/status/2016279378846724461)  2026-01-27T22:37Z [----] followers, [----] engagements


"And Trinity-Large-TrueBase is for those who love true base models. This is a 10T token checkpoint that has seen zero SFT data and no LR decay. It's magestic and very deep. It experienced one of the smoothest loss curves you can hope to see"  
[X Link](https://x.com/latkins/status/2016279380792901795)  2026-01-27T22:37Z [----] followers, 23.3K engagements


"@1thousandfaces_ @viemccoy @tautologer Were too poor after training to host the truebase as well. But Id consider sponsoring a week or so of node time if some people who truly get base models wanted to share some fun outputs"  
[X Link](https://x.com/latkins/status/2016294902695657569)  2026-01-27T23:39Z [----] followers, [----] engagements


"No way Apple has landed the rights to turn MISTBORN into a film franchise & THE STORMLIGHT ARCHIVE into a TV series. Brandon Sanderson will write produce and consult on all projects. (Source: https://t.co/Ka6RvxmT3S) https://t.co/b8FZH2ahYO Apple has landed the rights to turn MISTBORN into a film franchise & THE STORMLIGHT ARCHIVE into a TV series. Brandon Sanderson will write produce and consult on all projects. (Source: https://t.co/Ka6RvxmT3S) https://t.co/b8FZH2ahYO"  
[X Link](https://x.com/latkins/status/2016675487456879008)  2026-01-29T00:51Z [----] followers, [----] engagements


"Everyone likes to shit on grok but this IS very impressive Understanding requires imagining. Grok Imagine lets you bring whats in your brain to life and now its available via the worlds fastest and most powerful video API: https://t.co/tqQwQVgCEI Try it out and let your Imagination run wild. https://t.co/Bn6Z70Ual6 Understanding requires imagining. Grok Imagine lets you bring whats in your brain to life and now its available via the worlds fastest and most powerful video API: https://t.co/tqQwQVgCEI Try it out and let your Imagination run wild. https://t.co/Bn6Z70Ual6"  
[X Link](https://x.com/latkins/status/2016762182239998140)  2026-01-29T06:35Z [----] followers, 23K engagements


"@ericssunLeon They were pretty smart for that"  
[X Link](https://x.com/latkins/status/2016935861972287768)  2026-01-29T18:05Z [----] followers, [---] engagements


"@dhtikna @rasbt @grok we did some experiments with converting our dense gqa attention 4.5B model to KLA but @rasbt is dead on that we were too far into training prep to de-risk it at this scale: https://www.arcee.ai/blog/distilling-kimi-delta-attention-into-afm-4-5b-and-the-tool-we-used-to-do-it https://www.arcee.ai/blog/distilling-kimi-delta-attention-into-afm-4-5b-and-the-tool-we-used-to-do-it"  
[X Link](https://x.com/latkins/status/2016943956844793967)  2026-01-29T18:38Z [----] followers, [--] engagements


"Really smart but don't get it twisted this isn't to help you. This is to help them improve their swe models. People get paid for this kind of grading so just know what you're giving them for your time. Brilliant though tbh would do the same. Introducing Arena Mode in Windsurf: One prompt. Two models. Your vote. Benchmarks don't reflect real-world coding quality. The best model for you depends on your codebase and stack. So we made real-world coding the benchmark. Free for the next week. May the best model win. https://t.co/qXgd2K4Yf6 Introducing Arena Mode in Windsurf: One prompt. Two models."  
[X Link](https://x.com/anyuser/status/2017351495033962742)  2026-01-30T21:37Z [----] followers, [----] engagements


"@gazorp5 @windsurf If it isnt stated otherwise good chance it will at least be retained"  
[X Link](https://x.com/latkins/status/2017378762879541614)  2026-01-30T23:25Z [----] followers, [---] engagements


"I mean hell Id invest $10 BREAKING 🚨: Amazon is planning to invest from $10 to $60 billion in OpenAI according to The Information. These numbers just keep going up 👀 https://t.co/XvHyL5lKLa BREAKING 🚨: Amazon is planning to invest from $10 to $60 billion in OpenAI according to The Information. These numbers just keep going up 👀 https://t.co/XvHyL5lKLa"  
[X Link](https://x.com/anyuser/status/2017704260008612107)  2026-01-31T20:59Z [----] followers, 35.6K engagements


"@RicardoMonti9 I have a diversified portfolio"  
[X Link](https://x.com/latkins/status/2017822867078127878)  2026-02-01T04:50Z [----] followers, [---] engagements


"@thdxr Not if but when. And not just anthropic. All of them will eventually do this as the model is the product continues to become more prevalent"  
[X Link](https://x.com/latkins/status/2017827673536663877)  2026-02-01T05:09Z [----] followers, [----] engagements


"Yeah and presumably codex models are also highly optimized to do best inside codex. So even if made widely available their full potential is limited to the sandbox they know best. https://twitter.com/i/web/status/2017845922659434886 https://twitter.com/i/web/status/2017845922659434886"  
[X Link](https://x.com/latkins/status/2017845922659434886)  2026-02-01T06:22Z [----] followers, [---] engagements


"Today Arcee is releasing two datasets: [--]. The Tome - this is a [----] million sample dataset that has been filtered to train strong generalist models. This is the dataset that was used to train Spark and Nova [--]. Agent-Data: This is Arcee-Agent's dataset comprising different function calling datasets from salesforce internlm and glaive (with an extra 20k samples extended for multiple tool calls per response). This includes Magpie-300k-Pro as well to prevent overfitting and make the model a strong conversationalist. Enjoy Links below"  
[X Link](https://x.com/latkins/status/1815475969773445439)  2024-07-22T19:56Z [----] followers, [----] engagements


"@s_tworkowski Do you send reasoning traces via api"  
[X Link](https://x.com/latkins/status/1913316763250385393)  2025-04-18T19:40Z [----] followers, 19.8K engagements


"This is an insane opportunity btw. You likely wont get better experience outside of the big [--] (closed) labs. We're starting to hire for our [----] Olmo interns Looking for excellent students to do research to help build our best models (primarily enrolled in Ph.D. with experience or interest in any area of the language modeling pipeline). https://t.co/x1jA3WALk8 We're starting to hire for our [----] Olmo interns Looking for excellent students to do research to help build our best models (primarily enrolled in Ph.D. with experience or interest in any area of the language modeling pipeline)."  
[X Link](https://x.com/latkins/status/1986294167245627593)  2025-11-06T04:46Z [----] followers, 17K engagements

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