new name, same mission a young startup investing in a community-oriented research lab -- I never saw this, but I also never saw such a small team having such a disproportionate impact

Very proud of what we have achieved over last 3 years, and the breakthroughs ahead. 🔥 Our new name helps better communicate our work and our impact at the frontier of AI progress. Everything else stays the same, including our commitment to explore the unknown, together.

I’m so proud of all the work Cohere For AI has done over the past 3 years both as an industry lab and an active open science community that has given entry points to new researchers❣️ Super excited to continue doing the same and much more in our new chapter as Cohere Labs! 🚀

Honoured to have 2 of our datasets recognized by the Stanford HAI AI Index Report as some of the most significant releases of 2024. Congrats to the Aya Dataset & Global MMLU teams, with shoutout to Shivalika Singh - 1st author on both! - and mentors Marzieh Fadaee Beyza Ermiş Sara Hooker.

The impact of data is still wildly underappreciated — not just for multilingual models, but for every model 🔥 Proud to be part of projects that bring cutting-edge technology to previously under-represented languages 🌍

Super special to see both Aya Dataset and Global MMLU get recognised as part of Stanford HAI AI Index Report! 🥹 Both of these projects required a lot of care and intense collaboration! Shout out to all co-authors, contributors and mentors Sara Hooker Marzieh Fadaee Beyza Ermiş ! ❤️

I’m excited to share @Cohere’s newest model, Embed 4! Embed 4 is the optimal search engine for secure enterprise AI assistants and agents.

Embed 4 is natively multilingual across 100+ languages, especially excelling in the languages our customers speak, such as Arabic, French, Japanese, and Korean.

Today we are releasing Embed 4 – the new SOTA foundation for agentic enterprise search and retrieval applications! cohere.com/blog/embed-4 Check out the blog for similarly visually satisfying graphs :)

we did it again embed4 creates intelligent indices to help humans and EGI find needles in all the haystacks. SOTA across every dimension.

128K, multilingual, strong embedding model! Cohere is a very sane lab. They've picked their enterprise niche and do no-BS work there.

Cohere is the undisputed king of embeddings

Wow next gen embedding model: > 128k context length!! > Multilingual 100+ langs > Multimodal > Quantization aware training "rAG iS dEaD" fans in shambles.

Huge shoutout to cohere on the launch of Embed v4! Helping developers solve tough dynamic retrieval problems for agents and assistants, from complex documents (PDFs, y’all!) to multilingual search — we’re excited for what’s ahead!

Embed 4 is here! Best enterprise search model - 128K context length, natively multilingual and multimodal, right balance of accuracy with efficiency, and optimizations for key industries like finance, healthcare, and manufacturing. Excited to hear more customer feedback!

Best embedding model in the game by far

𝐂𝐨𝐡𝐞𝐫𝐞 𝐄𝐦𝐛𝐞𝐝 𝐯𝟒 - 𝐒𝐭𝐚𝐭𝐞-𝐨𝐟-𝐭𝐡𝐞-𝐚𝐫𝐭 𝐭𝐞𝐱𝐭 & 𝐢𝐦𝐚𝐠𝐞 𝐫𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥 Today we are releasing Embed v4, unlocking so many cool new features for retrieval. 🇺🇳 100+ languages 🖼️ Text & Image capabilities 📜 128k context length

Our latest models are now available on @GitHub Models! Try our latest generative model, Command A, and our brand new Embed 4, in the Github playground or API. github.blog/changelog/2025…

🚀🌍The rapid advancement of multilingual large language models (mLLMs) is exciting, but are we evaluating them effectively? Our new paper explores how we can improve generative evaluations for mLLMs by learning from machine translation (MT) evaluation practices. 🔎

🚨 Excited to share our latest paper! Multilingual LLMs are getting really good. But the way we evaluate them? Not the best sometimes. 🌟 We show how decades of lessons from Machine Translation can help us fix it