Farama-Foundation/Gymnasium

What is it?

Key features

• 01Standard reset/step API shared by all supported environments
• 02Built-in environment families including Classic Control, Box2D, Toy Text, MuJoCo and Atari
• 03Environments created by ID with `gymnasium.make`
• 04Optional dependency extras to install only the environment families you need
• 05Versioned environment IDs to keep results reproducible

Use cases

• Benchmarking a new reinforcement learning algorithm on standard tasks
• Teaching reinforcement learning with simple environments like CartPole and Toy Text
• Debugging agent implementations with small discrete environments
• Plugging a custom or third-party environment into existing agent code

Quick start

1. 1

   Install the base library with pip

   pip install gymnasium
2. 2

   Install extra dependencies for one environment family, such as Atari

   pip install "gymnasium[atari]"
3. 3

   Create an environment and step through it with the example code from the README

Things to know

• !Many environment families need extra dependencies, and some are difficult to install on certain systems.

Topics

FAQ

What is Gymnasium?

A standard Python API for single-agent reinforcement learning environments, plus a set of reference environments, forked from OpenAI's Gym. Gymnasium provides a common interface between reinforcement learning algorithms and the environments they train in. Each environment exposes reset and step methods along with observation and action spaces, so an agent written against one environment can be run on another without changes to the learning code.

Is Gymnasium free to use? What license does it have?

Gymnasium is open source under the MIT license. Check the license terms before using it in commercial products.

Is Gymnasium still maintained?

Actively maintained. The last commit to Gymnasium was on Oct 14, 2026.

What can I use Gymnasium for?

Benchmarking a new reinforcement learning algorithm on standard tasks Teaching reinforcement learning with simple environments like CartPole and Toy Text Debugging agent implementations with small discrete environments Plugging a custom or third-party environment into existing agent code

How do I install Gymnasium?

Install the base library with pip: pip install gymnasium Install extra dependencies for one environment family, such as Atari: pip install "gymnasium[atari]" Create an environment and step through it with the example code from the README

Where can I get Gymnasium?

The source code and installation instructions are on GitHub: https://github.com/Farama-Foundation/Gymnasium