Maia Chess

Maia Chess is a neural network chess engine trained to play like a human instead of finding the strongest move. It predicts the moves that real players of a given skill would choose, and can target any rating level on the Lichess scale from 600 to 2600. Because of that, it shares many of the mistakes and habits of human players, which makes it useful for realistic practice games and for understanding why people err. The site gathers several tools around the model. Users can play against Maia, analyze games with insights about what players at different rating levels tend to do in each position, work through curated training puzzles, and practice openings in a drill mode. It also offers Hand and Brain, a collaborative variant played with Maia, and Bot or Not, a game in which the player tries to tell human play from AI play. There is a broadcasts section as well. Sign in is through a Lichess account. It is aimed at club and improving players, coaches and researchers interested in human and AI collaboration. The model was introduced in a paper at KDD 2020, with Maia 2 described in a paper at NeurIPS 2024, both linked from the site. The site does not publish pricing or paid plans.