Gymnasium

Why we included this project

Gymnasium is the interface most people reach for when they need to train and compare reinforcement learning agents. It grew out of OpenAI's Gym, and the Farama Foundation now maintains it, so it has stayed compatible with the ecosystem that built up around the original while continuing to evolve. The library ships a solid collection of environments, from simple classic control problems and toy text tasks to physics-based Box2D games and MuJoCo simulations, along with wrappers and vectorized environments that handle the plumbing of common training loops. Because the same API sits under all of them, you can swap environments and compare algorithms without rewriting your code, which is exactly what makes it useful for both research and teaching.