google-deepmind/open_spiel

Health v2 · maintenance only

Adoption not part of the score

5435 stars · 1171 forks observed · 2026-08-28

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

OpenSpiel is a collection of game environments and reinforcement learning/search/planning algorithms for research in general RL and multiagent games. The core API and games are implemented in C++ with Python bindings, supporting a wide variety of game types and analysis tools.

Use cases

• research reinforcement learning in games
• implement multi-agent RL algorithms
• study game-theoretic learning dynamics
• benchmark planning and search algorithms on games
• model imperfect-information games like poker
• teach RL and game theory concepts

When to choose

• you need a standardized collection of game environments for RL research
• you want both C++ performance and Python usability
• you need support for many game classes (perfect/imperfect information, simultaneous moves, general-sum)
• you want to analyze learning dynamics across games

When to avoid

• you want to build a production game or game engine
• you need high-throughput 3D or physics-based environments
• you only need a simple single-agent RL gym without game-theoretic features

Facets

framework · maturity active

reinforcement-learning simulation game-engine machine-learning benchmarking reinforcement-learning artificial-intelligence windows cpp python multi-agent game-theory search-and-planning extensive-form-games research-framework game-development research algorithms linux macos

Member repositories

For agents

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