Olmo Hybrid is a fully open 7B model that combines transformer attention with linear RNN layers. Utilizing a 3:1 pattern of Gated DeltaNet to attention, it matches the accuracy of Olmo 3 on MMLU while using 49% fewer tokens.

SERA is a family of open coding models (8B, 14B, 32B) trained with a new efficient method. SERA learns from "soft-verified" data, drastically reducing training costs. Easily adaptable to private codebases. Open weights, data & recipes.

Molmo 2, a new suite of state-of-the-art vision-language models with open weights, training data, and training code, can analyze videos and multiple images at once.

AI2's flagship Olmo 3 model family empowers the open source community with not only state-of-the-art open models, but the entire model flow and full traceability back to training data.

Objaverse is A massive dataset with 800k+ annotated 3D objects. Objaverse improves upon present day 3D repositories in terms of scale, number of categories, and in the visual diversity of instances within a category.

Ai2 Paper Finder is the LLM-powered literature search that mimics the human research process. Uses iterative search and reasoning to find hard-to-find papers beyond keywords.