Ai2

Breakthrough AI to solve the world's biggest problems.

› Join us:  Newsletter:

[Ai2 @allen_ai]

Today we’re bringing new NSF OMAI compute online with NVIDIA Blackwell Ultra-powered systems, turning a $152M national investment from @NSF & @NVIDIA into a foundation for truly open AI research. 🧵

[Ai2 @allen_ai]

What can you build with a fully open robotics model in a weekend? 🤖

Robotics engineer @pham_blnh used MolmoAct 2, our open vision-language-action model, in the voice-controlled robot that won @southpkcommons' AI hackathon.

Watch our interview with him ↓ 🎥

[Ai2 @allen_ai]

Peer review is part of what makes ICML possible.

This year, at least 8 Ai2ers are contributing as area chairs or technical reviewers, including ICML-recognized Gold and Silver reviewers.

[Ai2 @allen_ai]

We're at #ICML2026 with papers & talks across the conference. Come say hello and learn about our latest research!

[Sewon Min @sewon__min]

Excited to see our FlexOlmo (https://t.co/NaQc8deXIc), released a year ago, is now being incorporated into the Danish Foundation Models (DFM) project!

[Ai2 @allen_ai]

Modular training is gaining momentum as frontier models become costlier to train & deploy.

This project shows how open, distributed approaches can make development more practical for national projects, public institutions, & smaller teams.

↓ Learn more: https://t.co/Wod1Erj1MM

[Ai2 @allen_ai]

The Danish Foundation Models (DFM) project is adapting our modular FlexOlmo architecture into a lighter-weight system that runs on commodity hardware—putting collaborative model building within reach of smaller research groups & organizations. 🧵

[Ai2 @allen_ai]

"FlexMoRE significantly reduces FlexOlmo's memory demands while preserving performance across almost all categories, allowing a broader audience to benefit from modular models," says Jacob Nielsen, who helped develop FlexMoRE at Ordbogen A/S and SDU's OdenseNLP lab.

[Ai2 @allen_ai]

We're at #ACL2026 with papers & talks across the conference. Come say hello and learn about our latest research!

[Ai2 @allen_ai]

@yoshitomo_cs Please DM with details and we'll look into it!

[Peter Sushko @PeterSushko]

DiScoFormer will be at ICML as an Oral. Big thanks to @allen_ai and @RanjayKrishna for support while working on this project!

[Ai2 @allen_ai]

Score & density show up across many fields. We hope one pretrained model like DiScoFormer can serve them all, at scale.

📝 Blog: https://t.co/bomZWcgcfy
📄 Report: https://t.co/g9M5oaYY2m

[Ai2 @allen_ai]

AI image generators don't "draw"—they follow a compass: the score function, which points toward more probable images. The same compass drives Bayesian sampling and plasma physics.

We built DiScoFormer to estimate the score far better when data gets complex. 🧵

[Ai2 @allen_ai]

DiScoFormer can also handle data unlike anything it trained on—more clusters than it ever saw, and unfamiliar shapes.

[Ai2 @allen_ai]

This update came directly from partners asking for cleaner embeddings. OlmoEarth v1.2 comes in Nano, Tiny, Small, & Base—all open source.

🤗 Models: https://t.co/9uGJLi9loK
💻 Training & fine-tuning code: https://t.co/QzncWyXuq6
📄 Tech report: https://t.co/bmeRA1zJwt

[Ai2 @allen_ai]

Today we're releasing OlmoEarth v1.2, the latest in our family of open foundation models for Earth observation. 🌍

We've switched to rotary positional embeddings (RoPE), which reduces artifacts in the embeddings & gives a small performance boost. 🧵

[Ai2 @allen_ai]

V1.2 switches to RoPE. Instead of adding a position signal to each patch, it rotates the vectors the model compares in attention by angles defined by each patch's position.

Across all model sizes, we see consistent improvement on our kNN/linear-probe evals.