SpinGPT: A Large-Language-Model Approach to Playing Poker Correctly
Published in Advances in Computer Games 2025, 2025
In this paper, we present SpinGPT, a poker AI built on top of a large language model.
SpinGPT is trained in two stages. First, we fine-tune it on 320,000 hands from my former professional poker history. Second, we apply reinforcement learning on 320,000 hands generated by a solver*. After these two phases, SpinGPT selects the game-theoretically best action roughly four times out of five and defeats Slumbot, the reigning poker AI world champion.
*A poker solver is a program that computes strategies close to the game-theoretic optimum in specific game configurations.
Recommended citation: Maugin, N. and Cazenave, T. (2025). "SpinGPT: A Large-Language-Model Approach to Playing Poker Correctly." Advances in Computer Games 2025.
Download Paper | Download Slides