Olmo 3.1 Think 32B

A reasoning model released with its weights, checkpoints, code, training data, and documented model flow, so researchers can inspect or change each stage.