Chessbook-1 outperforms Allie 2.0 and Maia-3 in human move prediction
October 19, 2026
Chessbook-1, a 199M parameter model, achieves a 96/100 score in predicting human moves and 81/100 in matching strength. It outperforms Maia-3 (79M) and Allie 2.0 (5.6B) in mimicking specific Lichess player styles and time control behaviors.
HOW THIS AFFECTS YOU
●
builderYou can use highly efficient, low-parameter models for specialized behavioral simulation tasks.
●
researcherSmaller parameter counts can effectively model complex human behavioral patterns in niche domains.