PyBullet
Bullet PhysicsPyBullet is the Python interface to the Bullet Physics SDK, a real-time collision-detection and rigid-body simulation library used for robotics, reinforcement learning and VR. It loads robots from URDF, ships example gym-style environments and robot models, and renders through a basic OpenGL viewer. The C++ SDK is maintained by Erwin Coumans in the bullet3 repository.
The repository's last push was in October 2025 and its newest GitHub release is from 2022, while PyPI carries a January 2025 build.
Openness
5 high confidence- license
- zlib(LICENSE.txt
- source
- public
- core features withheld
- no
The SDK is under the zlib license, which is OSI-approved and asks only that altered versions be marked as such. Bundled third-party code keeps its own terms, and no part of the product is held back. zlib is OSI-approved, so the license tier is osi.
- https://pypi.org/pypi/pybullet/json recorded 2026-09-26
PyPI metadata for pybullet: license zlib; classifier "License :: OSI Approved :: zlib/libpng License"; Homepage https://github.com/bulletphysics/bullet3.
- https://raw.githubusercontent.com/bulletphysics/bullet3/HEAD/LICENSE.txt recorded 2026-09-26
LICENSE.txt: the zlib license, applying to all files "except for the files under 'Extras' and examples/ThirdPartyLibs", which carry their own licenses.
- https://raw.githubusercontent.com/bulletphysics/bullet3/HEAD/README.md recorded 2026-09-26
README: no hosted or paid edition is mentioned.
- https://ungh.cc/repos/bulletphysics/bullet3/files/master recorded 2026-09-26
Full listing of the master branch; third-party LICENSE files sit only under Extras/ and examples/ThirdPartyLibs/; no enterprise, ee or commercial path.
Adoption
3 medium confidencePyPI installs of the pybullet package, whose metadata links to the bullet3 repository.
- https://pypi.org/pypi/pybullet/json recorded 2026-09-26
PyPI metadata for pybullet 3.2.7: Homepage https://github.com/bulletphysics/bullet3.
- https://pypistats.org/api/packages/pybullet/recent recorded 2026-09-26
pypistats: last_month 508,213 downloads for pybullet.
Capability
1 medium confidencePyBullet simulates rigid robots on the CPU and is easy to script, which made it a common first simulator for reinforcement learning. Its contact modeling is simpler than MuJoCo's, GPU support never left the experimental stage, and it renders only a basic viewer.
- https://raw.githubusercontent.com/bulletphysics/bullet3/HEAD/README.md recorded 2026-09-26
README: "real-time collision detection and multi-physics simulation for VR, games, visual effects, robotics, machine learning"; "Requirements for experimental OpenCL GPGPU support"; examples pybullet_envs.examples.enjoy_TF_AntBulletEnv_v0 and deep_mimic; no ray-tracing or throughput claim.
Verified 2026-09-26