Reasoning Gym is a community-created Python library that provides procedural dataset generators and algorithmically verifiable reasoning environments for training reasoning models with reinforcement learning, offering virtually infinite training data with adjustable complexity. It features over 100 tasks across domains such as algebra, arithmetic, computation, cognition, geometry, graph theory, and logic, with some tasks having a single correct answer and others accepting multiple valid solutions. The project can be installed via PyPI (pip install reasoning-gym) and includes a dataset gallery plus a quickstart to generate and verify tasks.
AI named Reasoning Gym in August 2026.
AI cited github.com in these answers, after arxiv.org.
github.com/open-thought/reasoning-gym 14%Other sites 86%