In our latest work, we ask “Can model merging help with task tradeoffs over models obtained from different training runs”? We extend model merging to a setup where you have many *generalist* LLM checkpoints showing performance tradeoffs.

Introducing Command R7B: the smallest, fastest, and final model in our R series of enterprise-focused LLMs! It delivers a powerful combination of state-of-the-art performance in its class and efficiency to lower the cost of building AI applications. cohere.com/blog/command-r…

Trying out Command R7B in VSCode, and the model performs brilliantly! 👏 The latest model from cohere Command family shows excellent performance working with code inside VSCode, using CodeGPT to integrate the model. Congrats to the Cohere team! 🥳 If you want to use Cohere's

mr. pretraining Acyr Locatelli is looking for intern to start in Jan, DM him if u can code

We are hiring cohere for an Agent Infrastructure Engineer! If you want to work on building the next generation of agent models for #RAG, #ToolUse #Code, #Reasoning and more then apply here. DM me if you have any Qs. jobs.ashbyhq.com/cohere/3f797fe…

Today cohere is very excited to introduce Command A, our new model succeeding Command R+. Command A is an open-weights 111B parameter model with a 256k context window focused on delivering great performance across agentic, multilingual, and coding usecases. 🧵

Today we are releasing Command A - Cohere’s newest model that offers enterprises powerful AI with minimum hardware :) It beats out bigger, slower models on enterprise agentic task performance, and can run on just two GPUs. Learn more about it: cohere.com/blog/command-a/

I'm excited to the tech report for our @Cohere Cohere For AI Command A and Command R7B models. We highlight our novel approach to model training including the use of self-refinement algorithms and model merging techniques at scale. Command A is an efficient, agent-optimised

Today (two weeks after model launch 🔥) we're releasing a technical report of how we made Command A and R7B 🚀! It has detailed breakdowns of our training process, and evaluations per capability (tools, multilingual, code, reasoning, safety, enterprise, long context)🧵 1/3.

Don't miss the detailed tech report for how we created Command A -- Cohere's flagship LLM!

Your next COBOL dev is a cohere model (Source: I made these tables)

We’re redefining what’s possible with AI. With the release of our latest model, Command A, optimized for real-world agentic and multilingual tasks, we’re demonstrating our commitment to bringing enterprises AI that goes beyond the ordinary, and offers security & efficiency.

Excited to finally share that CoPG — the RL method I co-authored with Nathan Grinsztajn and amazing colleagues — was used throughout the post-training (offline & online learning) of cohere’s new Command models! 🖊️ Tech report: cohere.com/research/paper… 🤖 CoPG: arxiv.org/abs/2406.19185

Preprint: Can we learn to reason for story generation (~100k tokens), without reward models? Yes! We introduce an RLVR-inspired reward paradigm VR-CLI that correlates with human judgements of quality on the 'novel' task of Next-Chapter Prediction. Paper: arxiv.org/abs/2503.22828

Command A, our state-of-the-art generative model, is now the highest-scoring generalist LLM on the Bird Bench leaderboard for SQL! It outperforms other systems that rely on extensive scaffolding to tackle these SQL benchmarks, and instead delivers these results out-of-the-box,