Ataraxos beats Stratego champion using 1/500th the compute

A new AI system defeated the most decorated human Stratego player of all time by 15 wins to 1 loss, with four draws, over a 20-game series, according to a paper published in Nature on 30 September 2026.

The system, Ataraxos, was built by researchers at Carnegie Mellon, MIT, NYU and Stanford. Training ran on 16 NVIDIA H100 GPUs for a week and cost 'a few thousand dollars,' which the authors say is roughly 1/500th the compute of the prior DeepNash effort. The paper calls the result 'to our knowledge, the first superhuman result in the game's history.'

The reference point is explicit. 'Even with multimillion-dollar industrial research efforts, top-human-level play at Stratego...has remained beyond the reach of artificial intelligence,' the abstract notes.

The same techniques produced a superhuman Barrage Stratego AI that defeated three multi-time world champions, state-of-the-art Hanabi scores across 2 to 5 player variants (24.654 with 77.53% perfect games in the 2-player setting), and a win over the previous dou dizhu benchmark, PerfectDou. The authors frame it as 'a design pattern for reinforcement learning and search that is effective under large amounts of hidden information.'

The headline 20-game series was against Pim Niemeijer. A separate 40-game exhibition at a world championship went 38 wins to 2 losses with no draws.