google-deepmind/lab

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

7373 stars · 1403 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

DeepMind Lab is a customisable 3D learning environment built on Quake III Arena (ioquake3) that provides navigation and puzzle-solving tasks for AI agents. It serves as a research testbed, primarily for deep reinforcement learning, with a Python API for agent training.

Use cases

• train deep reinforcement learning agents in 3D environments
• benchmark navigation and puzzle-solving capabilities of AI agents
• build custom 3D levels and tasks for agent-based AI research
• run a random agent demo to explore the environment API
• evaluate visual navigation and maze-solving in simulated 3D worlds

When to choose

• you need a challenging 3D testbed for deep RL research
• you want a customisable, open-source game-based learning environment
• you're reproducing or extending published DeepMind Lab experiments

When to avoid

• you need a modern, actively maintained RL environment
• you require Windows or macOS support (Linux-focused build)
• you want lightweight 2D or low-compute simulation

Facets

library · maturity maintenance

reinforcement-learning simulation game-engine machine-learning reinforcement-learning artificial-intelligence deep-learning simulation python cpp c 3d-environment agent-based-research quake3 bazel research-testbed game-development linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="google-deepmind/lab")