# LeelaChessZero/lc0

Open source neural network chess engine with GPU acceleration and broad hardware support.

Repository: https://github.com/LeelaChessZero/lc0
Canonical: https://ross.abutalabs.com/products/lc0
Homepage: https://lczero.org/
Language: C++
License: GPL-3.0
License Family: copyleft
Topics: alphazero-inspired, chess, chess-engine, leela-chess-zero, neural-networks, uci, alphazero, chess-ai, cuda, deep-learning, deep-reinforcement-learning, gpu
Last push: 2026-05-05T22:05:03+00:00

## Health v2 (maintenance only)
Score: 65/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 80, release rhythm 26, longevity 100
- inputs: {"age_days": 3017, "days_push": 120, "days_rel": 283, "gap_med": 199.0, "n_releases_24m": 3}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3193, forks 604 (observed 2026-08-28T04:07:48.247400+00:00)

## What it is
Lc0 is an open-source, UCI-compliant chess engine that plays chess using neural networks trained via AlphaZero-style self-play reinforcement learning. It supports GPU acceleration through CUDA, cuDNN, onnxruntime, Apple Metal, and experimental SYCL backends, plus CPU via OpenBLAS or DNNL.

## Use cases
- play chess against a neural network engine
- run a strong UCI chess engine in a chess GUI
- analyze chess positions with deep positional insight
- generate self-play games for neural network training
- benchmark GPU inference for board game AI
- study AlphaZero-style reinforcement learning in practice

## When to choose
- you want a strong, human-bias-free neural network chess opponent or analysis engine
- you have a GPU (NVIDIA, AMD, Intel, or Apple) and want accelerated engine play
- you want to contribute to or study open AlphaZero-style self-play training
- you need a UCI engine compatible with standard chess GUIs

## When to avoid
- you need a lightweight CPU-only engine with minimal dependencies
- you want a traditional hand-crafted evaluation engine like Stockfish's classical approach
- you need a general-purpose game AI framework rather than a chess-specific engine
- you cannot install GPU toolchains like CUDA or onnxruntime and need top-level strength

## Facets
- artifact type: application
- maturity: active
- function: machine-learning, deep-learning, reinforcement-learning, gpu-computing, game
- domain: machine-learning, deep-learning, reinforcement-learning, gpu-computing
- platform: windows, cpp, cli
- tags: chess-engine, uci, alphazero, neural-network, self-play, mcts, chess-ai, onnxruntime, metal, game-development, linux, macos, cuda, gpu

## Member repositories
- LeelaChessZero/lc0 (main) score 65

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:48.247400+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:24:54.992759+00:00, confidence not recorded.
  - readme: https://github.com/LeelaChessZero/lc0 (fetched 2026-08-28T04:07:48.247400+00:00, sha 593c058c8c01)
  - homepage: https://lczero.org/ (fetched 2026-08-29T09:38:40.552099+00:00, sha 0b14eeb371f3)
  - site_page: https://lczero.org/play/quickstart (fetched 2026-08-29T09:38:40.554499+00:00, sha 853ad0f08f14)
- Data as of 2026-08-30T08:39:29.467469+00:00.