Open models impact
Fully open models create opportunities to study AI systems in ways closed systems cannot. Across Olmo, Molmo, and our robotics work, we release models and research artifacts that let others look beneath the surface—examining how capabilities emerge, testing how systems behave and why, and building new methods on top of them.
September 2026 - How Goodfire used Ai2’s open post-training stack to trace unwanted model behavior
Goodfire used Ai2’s fully open post-training stack to predict LLM behavioral changes, trace unwanted model behavior...
August 2026 - How researchers adapted Dolma for better Thai language models
Thai researchers adapted Ai2’s open Dolma toolkit to build Mangosteen, a 47-billion-token Thai corpus that filters...
August 2026 - How a Georgia Tech team used the open Olmo stack to trace social reasoning
A Georgia Tech team used Ai2’s fully open Olmo stack to trace social reasoning back to the training data that...
August 2026 - When a model reads a drug's class from its name—not its knowledge
Researchers used Olmo 3 and its open training data to show that models can infer a drug’s class from its name instead...
July 2026 - MolmoAct 2 shows what open models can unlock for robotics
Robotics engineer Binh Pham used MolmoAct 2 to build a voice-controlled robot that won South Park Commons’…
July 2026 - Tracing distinctive language in AI-written text
Stony Brook researchers used our infini-gram engine to trace distinctive phrases in AI-generated writing back to…
July 2026 - Modular LLMs at scale: how FlexOlmo is helping to pool national expertise without pooling sensitive data
Danish Foundation Models is using FlexOlmo as the basis for FlexMoRE, a more efficient modular LLM architecture…
June 2026 - How Domyn and AISquared built on Ai2's open releases
Domyn and AISquared show how Ai2’s open releases are helping AI labs build models for regulated industries, where…
May 2026 - Building accessibility tools on a truly open foundation
PointCheck, an independent project, uses Molmo, MolmoWeb, and Olmo 3 to test web accessibility the way a keyboard…
May 2026 - Why Artificial Analysis uses Ai2's IFBench instruction-following eval
Artificial Analysis uses Ai2’s open IFBench eval because it captures a stubborn, real-world capability many…
October 2025 - Studying how models learn
A KAIST AI study shows how Olmo’s open checkpoints and training data let researchers observe, step by step, how…
October 2025 - Machine unlearning with Olmo
A Saarland University and Mila/McGill study uses Olmo’s open weights and Dolma training data to show why machine…
October 2025 - Tracing knowledge cutoffs
A Johns Hopkins–led study uses Olmo’s open training data, code, and model history to show that an LLM’s real…
October 2025 - Watching an LLM learn math skills
A Stanford-led MathCAMPS study uses Olmo’s open intermediate checkpoints to trace when specific math skills…
October 2025 - Clinical NLP using Olmo
A Northeastern University study uses Olmo’s open weights and accessible activations to identify where demographic…
October 2025 - Olmo and equivalent facts
An Oxford-led study uses Olmo’s open training data and checkpoints to show that language models can treat…