Introducing ✨ Aya Vision ✨ - an open-weights model to connect our world through language and vision Aya Vision adds breakthrough multimodal capabilities to our state-of-the-art multilingual 8B and 32B models. 🌿

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. 🧵

Challenge and success, indeed Command A, indeed.

This is what we've been hard at work for the last few months :) Command A is great at long context (256k easily), multilinguality, and throughput overall. Pre-training the base model and all the work leading up to that was super rewarding. I'm very happy it's out now 😌

🚀 Big news cohere's latest Command A now climbs to #13 on Arena! Another organization joining the top-15 club - congrats to the Cohere team! Highlights: - open-weight model (111B) - 256K context window - $2.5/$10 input/output MTok More analysis👇

shoutout taylor swift

📄 You can find the full tech report at cohere.com/research/paper…

push it to the [infinite width] limit youtube.com/watch?v=pP6ee4…

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.

This year is really full of muP!!

1/ Science is only as strong as the benchmarks it relies on. So how fair—and scientifically rigorous—is today’s most widely used evaluation benchmark? We took a deep dive into Chatbot Arena to find out. 🧵

I'm excited to share our new pre-print ShiQ: Bringing back Bellman to LLMs! arxiv.org/abs/2505.11081 In this work, we propose a new, Q-learning inspired RL algorithm for finetuning LLMs 🎉 (1/n)

Fun sonnet 4 hallucination on muP The Yang-Lecun correspondence

🔠 UTF-8 was never meant for language models. Yet every major tokenizer still uses it, creating unfair "byte premiums". Why should your native script cost more to tokenize? It's time for a change. 🧵👇

6/6 👇 Full paper & code 👇 📄 arxiv.org/abs/2505.24689

How can we make language models more flexible to adapt to new languages after pretraining? 🌏 🧠 Our latest work investigates whether a tokenizer trained on more languages than the pretraining target can improve language plasticity without compromising pretraining performance.

A huge thank you to all of my mentors and collaborators, especially Ahmet Üstün, Sara Hooker, Alejandro, and Marzieh Fadaee for their guidance and support ✨ 📜Check out our paper! arxiv.org/abs/2506.10766

This is very cool. One of the reasons I think muP hasn't caught on is that it is not seamlessly integrated with torch. Optax can make some things annoying, but this one is nice :)