Profile avatar
ai2.bsky.social
Breakthrough AI to solve the world's biggest problems. › Join us: http://allenai.org/careers › Get our newsletter: https://share.hsforms.com/1uJkWs5aDRHWhiky3aHooIg3ioxm
43 posts 2,987 followers 101 following
Regular Contributor
Active Commenter
Posts 15 Comments 29

Ali Farhadi joined the @geekwire.com podcast with @toddbishop.bsky.social to talk about: 🧠 The state of AI industry post-DeepSeek 📈 The building momentum around truly open AI 📲 Why on-device AI was a priority with OLMoE 💡 How we will use AI to solve real problems in 2025 Listen: buff.ly/41kjuoR

submitted 6 days ago • 0 comments

We took our most efficient model and made an open-source iOS app📱but why? As phones get faster, more AI will happen on device. With OLMoE, researchers, developers, and users can get a feel for this future: fully private LLMs, available anytime. Learn more from @soldaini.net👇 youtu.be/rEK_FZE5rqQ

submitted 13 days ago • 2 comments

ICYMI – we released a new model today, proving yet again that the performance gap between open vs. closed models is shrinking. Meet Tülu 3 405B.

submitted 25 days ago • 0 comments

Here is Tülu 3 405B 🐫 our open-source post-training model that surpasses the performance of DeepSeek-V3! It demonstrates that our recipe, which includes RVLR scales to 405B - with performance on par with GPT-4o, & surpassing prior open-weight post-trained models of the same size including Llama 3.1.

submitted 25 days ago • 2 comments

Only a few more days to submit your AI + Scientific Discovery papers to the AISD workshop @naaclmeeting.bsky.social ! The cross-submission form is now active -- have a submitted or already-published paper and want to present it to scientific discovery colleagues? Submit today!

submitted 27 days ago • 1 comment

Can AI really help with literature reviews? 🧐 Meet Ai2 ScholarQA, an experimental solution that allows you to ask questions that require multiple scientific papers to answer. It gives more in-depth and contextual answers with table comparisons and expandable sections 💡 Try it now: scholarqa.allen.ai

submitted 34 days ago • 1 comment

Some of our amazing teammates are organizing this exciting workshop for NAACL 2025 — submit your papers by January 30! 🎉

submitted 40 days ago • 0 comments

✍️Reminder: we've got open positions on our OLMo team for predoctoral candidates! Apply to be a Predoctoral Young Investigator by January 15 — job-boards.greenhouse.io/thealleninst...

submitted 48 days ago • 0 comments

Buckle your seatbelt — we've released the OLMo 2 paper to kick off 2025 🔥. Including 50+ pages on 4 crucial components of the LLM development pipeline.

submitted 49 days ago • 1 comment

New year, new channel — we're on Discord! Come join the conversation: discord.gg/NE5xPufNwu

submitted 49 days ago • 0 comments

Last year, we developed ACE, a boundary-breaking climate emulator. Now, we introduce ACE2, which addresses stability and accuracy across a range of climates, broadening the scope of how AI can help us deal with climate change. Learn more: allenai.org/blog/ai2-cli...

submitted 76 days ago • 0 comments

Remember Molmo? The full recipe is finally out! Training code, data, and everything you need to reproduce our models. Oh, and we have updated our tech report too! Links in thread 👇

submitted 77 days ago • 1 comment

We partnered with researchers @ucsandiego.bsky.social to supercharge climate modeling, speeding up predictions and enhancing accuracy.🌍 Check out the blog post below and come find us at #NeurIPS this week!

submitted 77 days ago • 1 comment

Meet OLMo 2, the best fully open language model to date, including a family of 7B and 13B models trained up to 5T tokens. OLMo 2 outperforms other fully open models and competes with open-weight models like Llama 3.1 8B — As always, we released our data, code, recipes and more 🎁

submitted 90 days ago • 5 comments

Meet Tülu 3, a set of state-of-the-art instruct models with fully open data, eval code, and training algorithms. We invented new methods for fine-tuning language models with RL and built upon best practices to scale synthetic instruction and preference data. Demo, GitHub, paper, and models 👇

submitted 95 days ago • 3 comments