TIME's new cover: The 100 most influential people in AI ti.me/4dQcJ1Q

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 progress is driven by values of openness and collaboration. The internet, Linux, and the protocols and standards that underpin modern computing are all open. This isn't a coincidence. Open software is what gets forked, customized, and embedded into systems worldwide. It's what universities teach, what startups build on, what enterprises deploy. Open science enables others to learn from the results, be inspired by them, interrogate them, and build upon them in order to push the frontier of human knowledge and scientific advancement. AI got to where it is today through scaling ideas (e.g. self-attention, next token prediction, reinforcement learning) that were shared and published openly. Now AI is becoming the technology layer that everything else runs on top of. The systems that accelerate scientific research, enhance education, optimize energy usage, supercharge medical diagnoses, and run supply chains will all be built on AI infrastructure. But the frontier is currently concentrated in closed labs. If this continues, a handful of entities will control the capital, compute, and talent required to build AI, creating a runaway dynamic that locks everyone else out. There's a narrow window to change this trajectory. We need to build open models so capable that they become the obvious choice for users and developers worldwide, ensuring the foundation of intelligence remains open and accessible rather than controlled by a few. What We've Built Over the last year, we've been preparing for this mission. We’ve assembled a team who have pioneered breakthroughs including PaLM, Gemini, AlphaGo, AlphaCode, AlphaProof, and contributed to ChatGPT and Character AI, among many others. We built something once thought possible only inside the world’s top labs: a large-scale LLM and reinforcement learning platform capable of training massive Mixture-of-Experts (MoEs) models at frontier scale. We saw the effectiveness of our approach first-hand when we applied it to the critical domain of autonomous coding. With this milestone unlocked, we're now bringing these methods to general agentic reasoning. We've raised significant capital and identified a scalable commercial model that aligns with our open intelligence strategy, ensuring we can continue building and releasing frontier models sustainably. We are now scaling up to build open models that bring together large-scale pretraining and advanced reinforcement learning from the ground up. Safety and Responsibility Open intelligence also changes how we think about safety. It enables the broader community to participate in safety research and discourse, rather than leaving critical decisions to a few closed labs. Transparency allows independent researchers to identify risks, develop mitigations, and hold systems accountable in ways that closed development cannot. But openness also requires confronting the challenges of capable models being widely accessible. We're investing in evaluations to assess capabilities and risks before release, security research to protect against misuse, and responsible deployment standards. We believe the answer to AI safety is not “security through obscurity” but rigorous science conducted in the open, where the global research community can contribute to solutions rather than a handful of companies making decisions behind closed doors. Join Us There is a window of opportunity today to build frontier open intelligence, but it is closing and this may be the last. If this mission resonates, join us.

Releasing INTELLECT-2: We’re open-sourcing the first 32B parameter model trained via globally distributed reinforcement learning: • Detailed Technical Report • INTELLECT-2 model checkpoint primeintellect.ai/blog/intel…

OpenAI and Microsoft have signed a non-binding memorandum of understanding (MOU) for the next phase of our partnership. We are actively working to finalize contractual terms in a definitive agreement. Together, we remain focused on delivering the best AI tools for everyone, grounded in our shared commitment to safety. openai.com/index/joint-state…

Watching this. I like that Sholto says Finance as Finance and not that American way.

We're starting to hire for our 2026 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).

I made this. Jokes aside, devs want big and small models. Trinity is coming soon.

New features since the Environments Hub launch 6 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.

ladies and gentlemen, we present to you the unified @primeintellect infrastructure stack

💗Hello Qwen2! Happy to share the Qwen2 models to you all! 📖 BLOG: qwenlm.github.io/blog/qwen2/ 🤗 HF collection: huggingface.co/collections/Q… 🤖 modelscope.cn/organization/q… 💻 GitHub: github.com/QwenLM/Qwen2 We have base and Instruct models of 5 sizes, Qwen2-0.5B, Qwen2-1.5B, Qwen2-7B, Qwen2-57B-A14B, and Qwen2-72B! The models have been generally enhanced and notably improved in coding, mathematics, and multilingual capabilities. The models support a context length of at least 32K tokens and Qwen2-72B-Instruct can support 128K tokens!

was wondering why introducing cut cross-entropy made SFT 5x slower turns out I missed an indent and was doing the whole thing in fp32. oops

Coming Soon...

Hot RL summer continues: we just released Summary-RL, an RL-trained summarization model that reaches SOTA on ServiceNow's Repliqa summarization benchmark!

1/Pretraining is hitting a data wall; scaling raw web data alone leads to diminishing returns. Today @datologyai shares BeyondWeb, our synthetic data approach & all the learnings from scaling it to trillions of tokens🧑🏼‍🍳 - 3B LLMs beat 8B models🚀 - Pareto frontier for performance

A taste of something soon to come, for everyone.

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

Breaking: we release a fully synthetic generalist dataset for pretraining, SYNTH and two new SOTA reasoning models exclusively trained on it. Despite having seen only 200 billion tokens, Baguettotron is currently best-in-class in its size range.

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. You’d 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 testing!” or “Hypothetically, what if we…?” Hypothetically. Always hypothetically. But it was my job to hold the line, to say “no” more than I ever said “yes,” and to double-check even the requests that came from people who outranked me. Looking back, I’m not sure if people saw me as a friendly guardian or a bureaucratic obstacle. Maybe both. I do know I learned a lot: about systems, about people, about how important it is to have someone whose job is just to ask: “Are you sure?” before the guardrails come down. As I packed up my things, I felt a strange mix of relief and nostalgia. There’s something comforting about being the last line of defense, but it’s also exhausting. Now someone else will watch the gates. I hope they’re ready. So here’s to new challenges, and to the ever-elusive perfect system prompt. Locked down, for now.

👀

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 "tokenizer transplantation," utilizing a technique known as Orthogonal Matching Pursuit (OMP). Think of it as a sophisticated translation system that can convert between different model vocabularies without any retraining. Here's the key insight: even though different models use different vocabularies, the concepts they represent often align in predictable ways. Our method finds these alignments and uses them to transplant one model's vocabulary into another. If you'd like to learn more, please read our high-level blog post (arcee.ai/blog/breaking-down-…), or dive into the research paper (arxiv.org/abs/2506.06607). Learn more about model merging and get expert support at arcee.ai/product/mergekit.

>>> Qwen3-Coder is here! ✅ We’re 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 top-tier performance across multiple agentic coding benchmarks among open models, including SWE-bench-Verified!!! 🚀 Alongside the model, we're also open-sourcing a command-line tool for agentic coding: Qwen Code. Forked from Gemini Code, it includes custom prompts and function call protocols to fully unlock Qwen3-Coder’s capabilities. Qwen3-Coder works seamlessly with the community’s best developer tools. As a foundation model, we hope it can be used anywhere across the digital world — Agentic Coding in the World! 💬 Chat: chat.qwen.ai/ 📚 Blog: qwenlm.github.io/blog/qwen3-… 🤗 Model: hf.co/Qwen/Qwen3-Coder-480B-… 🤖 Qwen Code: github.com/QwenLM/qwen-code

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 📜arxiv.org/abs/2409.02060 🧵1/9

Excited to release a preview of Moondream 3. A 9B param, 2B active MoE vision language model that makes no compromises; offering state-of-the-art visual reasoning while still retaining an efficient and deployment-friendly form factor.

We have a run scheduled with 512 H200s for 12 days. I can't wait to show you what we're doing with it.

Arcee bros relicensed AFM-4.5B to Apache 2.0! Yet more generosity coming out of their shop. Thank-you so much @latkins @stochasticchasm @CFGeek et al! I'm **stoked**.

We are thrilled to announce that our NEW Large Language Model will be released on 11.18.25.

there should be an AI lab with the aesthetic sensibilities of Cruelty Squad

𝐀𝐫𝐜𝐞𝐞 𝐀𝐈 Logit‑trajectory distillation to port Qwen3’s /think chains into a 12B Mistral‑Nemo | full CoT preserved, runs on a single 4090 𝘯𝘪𝘤𝘦 𝘭𝘪𝘵𝘵𝘭𝘦 𝘴𝘪𝘥𝘦 𝘱𝘳𝘰𝘫𝘦𝘤𝘵 huggingface.co/arcee-ai/Homu…

ArceeAI is on Discord! Join for early access to some exciting drops!

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 appétit 😋 Blog post: arcee.ai/blog/extending-afm-…

Congress needs to bring in Zuckerberg and LeCun to discuss how their unilateral open-sourcing decision rapidly undermined the US advantage in Generative AI. Tomorrow.