Leela Chess Zero
Neural-network chess engine trained via self-play. Strategic, human-like play that often disagrees with Stockfish.
No affiliate relationship with this tool. The link below goes directly to the product site.
Our take
Leela is my go-to when Stockfish and I disagree about a positional judgement call. Getting a second, genuinely different opinion from an engine trained on self-play rather than handcrafted evaluation has changed how I think about a few middlegame structures. The GPU requirement is the real friction — if you don't have one, you're better off just using Stockfish.
Leela Chess Zero (Lc0) is a neural-network chess engine inspired by DeepMind’s AlphaZero. Unlike Stockfish, which uses a handcrafted evaluation plus search, Lc0 learns entirely from self-play games and uses a deep neural network to evaluate positions.
The community publishes progressively stronger networks; users download a net and run it under their preferred chess GUI.
Pros
- Plays noticeably more "human" and positional than classical engines — great second opinion
- Completely free and open source (GPL v3), community-trained networks published regularly
- Often preferred for closed, strategic positions where Stockfish's brute force is less effective
Cons
- Requires a capable GPU to run at full strength — CPU-only use is painful
- Setup is more involved than Stockfish; you pick a network, a backend, a GUI
- Not always stronger than Stockfish in sharp tactical positions
Related tools
ChessBase
The professional desktop software for database work, deep analysis, and tournament preparation. Includes the Fritz engine.
Chessify
Cloud engine analysis in a browser. Run Stockfish at massive server speeds without installing anything.
Lichess Analysis Board
Free, open-source, ad-free analysis with Stockfish in the browser and a serverless cloud engine for deeper dives.