---
title: "Google DeepMind's MuJoCo Menagerie Gives Robotics Researchers 40+ Ready-to-Simulate Robots"
subtitle: "Google DeepMind's curated library of physics-ready robot models now spans 60+ humanoids, quadrupeds, arms and grippers with GPU-friendly MJX variants."
url: "https://alphasignal.ai/news/google-deepmind-s-mujoco-menagerie-gives-robotics-researchers-40-ready-to"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "Google DeepMind"
company_url: "https://alphasignal.ai/company/googledeepmind"
published: "2026-05-07T18:12:12.000Z"
updated: "2026-08-22T03:53:25.588Z"
type: "Repo"
topics:
  - "Robotics"
  - "Open Source"
subtopics:
  - "Vla Models"
  - "Humanoid"
  - "Manipulation"
upvotes: 45
source_url: "https://github.com/google-deepmind/mujoco_menagerie"
comments: 0
truncated: true
---

# Google DeepMind's MuJoCo Menagerie Gives Robotics Researchers 40+ Ready-to-Simulate Robots

> Google DeepMind's curated library of physics-ready robot models now spans 60+ humanoids, quadrupeds, arms and grippers with GPU-friendly MJX variants.

Building a physics simulation for robotics learning usually starts with a frustrating detour: hunting down a URDF, converting it to MJCF, wrestling with mass and inertia values, and discovering after two hours that your quadruped explodes on contact. [MuJoCo Menagerie](https://github.com/google-deepmind/mujoco%5Fmenagerie) exists to skip that ritual. It is a curated repository of ready-to-simulate robot models maintained by Google DeepMind, and it has quietly become the default starting point for anyone doing reinforcement learning, sim-to-real, or controls research on top of MuJoCo.

A physics simulator is only as good as the model it runs, and MuJoCo's flexibility makes it easy to author files that look right but behave badly. Menagerie's pitch is that every model in the collection has already been tuned to load, stand, and move without immediately falling through the floor. No mystery inertia tensors, no unstable contact geometries.

## What is actually in the box

The repository covers a broad slice of modern robotics hardware, organized by category. Some highlights:

* **Humanoids:** Unitree G1 (29 DoF) and H1 (19 DoF), Apptronik Apollo (32 DoF), PAL TALOS (44 DoF), Booster T1, Fourier N1, Berkeley Humanoid, Robotis OP3, and the ToddlerBot 2XC/2XM.

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [google-deepmind/mujoco_menagerie](https://github.com/google-deepmind/mujoco_menagerie)
- [mujoco_menagerie/google_robot](https://github.com/google-deepmind/mujoco_menagerie/tree/main/google_robot)
- [google-deepmind/mujoco_menagerie | DeepWiki](https://deepwiki.com/google-deepmind/mujoco_menagerie)
- [MuJoCo Model Gallery Documentation](https://mujoco.readthedocs.io/en/latest/models.html)
- [robot_descriptions.py](https://github.com/robot-descriptions/robot_descriptions.py)