Skip to content
RobothecaBeta
Log in

Brax

  • Software
  • Simulators
  • Simulator

Brax is a fast, differentiable physics engine written in JAX, used for robotics, reinforcement learning and other simulation-heavy research. It includes built-in training algorithms.

Made by Google, United States (checked on github.com/google (opens in a new tab)).

Source on GitHub: github.com/google/brax (opens in a new tab). Facts read at commit 6d03937 (opens in a new tab).

Licenses

Code
Apache-2.0 (opens in a new tab)

Read by Robotheca from LICENSE (opens in a new tab) (Apache-2.0) at commit 6d03937 (opens in a new tab), on 16 Oct 2026.

Commercial use
Yes

What it does

  • Simulates environments at millions of physics steps per second on TPU and scales to massively parallel simulation across devices.
  • Offers four interchangeable physics pipelines: MuJoCo XLA (MJX), Generalized, Positional and Spring, sharing one API.
  • Includes learning algorithms such as PPO, SAC, ARS, evolutionary strategies and analytic policy gradients that exploit differentiability.
  • Can be installed from PyPI, Conda or source, and provides a `learn` command to train a model.
  • Supports training on NVIDIA GPUs once CUDA, cuDNN and GPU-enabled JAX are installed.

Written by Robotheca from the project's README (opens in a new tab) at commit 6d03937.

Getting started

Difficulty
Advanced
Hardware it needs, as the project states it
Runs on single devices and scales to multiple devices; designed for acceleration hardware (TPU is mentioned for simulation speed). NVIDIA GPU training is supported after installing CUDA, cuDNN and JAX with GPU support. No specific memory or OS requirements are stated.

Maintenance

Status
Active: a commit in the last 90 days
Newest commit
7 Oct 2026 on main
Latest release
v0.14.2 (15 Mar 2026)

Checked on GitHub on 16 Oct 2026.

Links

Robots on Robotheca that use it

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists software as a reference: nothing is hosted or run here, and every fact links to where it was read.