Models
Models from the labs in Epoch AI's catalog, the repositories they published on Hugging Face, and posts from their own news pages.
Can we trust LLMs to judge whether a research idea is novel?Allen Institute for AIOct 27LitPivot: Developing research ideas through dynamic contextualization within the literature landscapeAllen Institute for AIOct 23A crop-mapping benchmark for the tropicsAllen Institute for AIOct 22OlmoEarth v1.3: State-of-the-art embeddings for the whole planetAllen Institute for AIOct 22FairStress: A deeper evaluation of fairness in LLM decision-makingAllen Institute for AIOct 20HiRO-ACEv2: Simulating extreme weather at 3 km resolution for a fraction of the costAllen Institute for AIOct 14How Molmo 2’s open release helped an MIT researcher rethink fish trackingAllen Institute for AIOct 13Ai2 at COLM 2026: Open research, from models to agents for scienceAllen Institute for AIOct 9Impactful scheduling for GPU clustersAllen Institute for AIOct 9Now in Nature: Retrofitting language models to operate over bytesAllen Institute for AIOct 7Open-sourcing AstaBrief, the fast report-generation model in AstaAllen Institute for AIOct 2Introducing Olmo-core 3: Open, scalable training infrastructure for large MoEsAllen Institute for AIOct 1What a crowdsourced game revealed about steering Olmo 3Allen Institute for AISep 17Teaching future scientists to interrogate AI tools for scientific discoveryAllen Institute for AISep 14How Goodfire used Ai2’s open post-training stack to trace unwanted model behaviorAllen Institute for AISep 9The hard parts of AI-assisted scienceAllen Institute for AISep 1BenchMIRT: What are LLM benchmarks actually measuring?Allen Institute for AISep 1Ai2 and Providence Swedish Cancer Institute partner to advance AI-assisted scientific discoveryAllen Institute for AIAug 27How researchers adapted Dolma for better Thai language modelsAllen Institute for AIAug 26How a Georgia Tech team used the open Olmo stack to trace social reasoningAllen Institute for AIAug 21When a model reads a drug's class from its name—not its knowledgeAllen Institute for AIAug 18TutorMoments: Do AI tutors know when to help and when to hold back?Allen Institute for AIAug 7Ai2 expands collaboration with Hugging Face to accelerate open scienceAllen Institute for AIAug 6Tracing distinctive language in AI-written textAllen Institute for AIJul 31The OlmoEarth Platform: Geospatial inference at planetary scaleAllen Institute for AIJul 28