Olmo 3.1 is an open-source language model from the Allen Institute for AI (Ai2) with extended reinforcement learning for stronger reasoning and benchmark performance.