A standard Python API for single-agent reinforcement learning environments with a collection of reference environments and utilities.
Gymnasium is an open-source Python library that provides a standard API for single-agent reinforcement learning environments. It enables developers and researchers to develop, compare, and benchmark reinforcement learning algorithms by offering a consistent interface and a collection of reference environments. The project is the maintained fork of OpenAI's Gym, ensuring ongoing support and evolution.
Reinforcement learning researchers, AI practitioners, and developers who need a reliable, standardized environment interface for training and evaluating agents.
Developers choose Gymnasium for its robust, community-maintained API, diverse environment families, and strict versioning that guarantees reproducibility, making it the de facto standard for reinforcement learning experimentation.
An API standard for single-agent reinforcement learning environments, with popular reference environments and related utilities (formerly Gym)
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Provides a unified interface for environment interaction, enabling algorithm portability and direct comparison across studies, as emphasized in the documentation for reproducible research.
Includes multiple families like Classic Control, Box2D, and Atari, offering varied complexity for testing and benchmarking agents, with third-party compatibility for extended use.
As the official successor to OpenAI Gym, it is actively maintained by the Farama Foundation with ongoing updates, community contributions, and a public discord for coordination.
Environment versions (e.g., '-v0') ensure that changes don't break existing experiments, crucial for academic benchmarking and consistent results.
The library does not officially support Windows; while PRs are accepted, users may encounter installation and runtime issues on this platform, as noted in the README.
Installing all dependencies with 'gymnasium[all]' can be massive and problematic, especially for environments with complex setups like MuJoCo, leading to potential system conflicts.
Gymnasium focuses solely on environments, requiring integration with separate libraries like CleanRL for agent training, which adds complexity to setting up full reinforcement learning pipelines.