leela-zero/leela-zero

Health v2 · maintenance only

Adoption not part of the score

5586 stars · 1012 forks observed · 2026-08-28

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

Leela Zero is an open-source Go engine that reimplements AlphaGo Zero, combining Monte Carlo Tree Search with a deep residual convolutional neural network and requiring no human-provided game knowledge. It relies on externally supplied network weights and supports a distributed community effort to train them.

Use cases

• play Go against a strong AI engine
• reproduce the AlphaGo Zero paper
• contribute GPU compute to a distributed Go training effort
• study reinforcement learning and MCTS in board games
• run a self-taught Go program without human game records

When to choose

• you want an open-source AlphaGo Zero-style Go engine
• you have a modern GPU and want to play or help train the network
• you want to study MCTS plus deep neural network game play

When to avoid

• you need a strong engine out of the box without obtaining network weights
• you lack a GPU and need high performance
• you want a Go engine trained on human professional games

Facets

application · maturity maintenance

machine-learning deep-learning gpu-computing simulation artificial-intelligence deep-learning reinforcement-learning windows cpp go-engine alphago-zero mcts reinforcement-learning neural-network board-games distributed-computing game-development linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="leela-zero/leela-zero")