GPT-6.1 Sol Medium
Answered each puzzle once; the other models answered each 3 times. Fewer samples make this score noisier: a rerun could land several points higher or lower, so read its rank as approximate.
42.2Overall
By game
GameQualityLegalityCostnRank
Xiangqi33.3100.0$0.0332 of 4
xiangqi-01g6g40.0100.0$0.03
xiangqi-02b4a60.0100.0$0.04
xiangqi-03h4h0100.0100.0$0.03
Chess0.0100.0$0.0234 of 4
chess-01c6c30.0100.0$0.03
chess-02c6c50.0100.0$0.02
chess-03g8f80.0100.0$0.02
Go34.4100.0$0.0233 of 4
go-01C6100.0100.0$0.01
go-02P61.4100.0$0.02
go-03E111.9100.0$0.02
Backgammon56.9100.0$0.0132 of 4
backgammon-017/1 5/40.0100.0$0.01
backgammon-0213/770.7100.0$0.01
backgammon-0313/4100.0100.0$0.01
Shogi73.8100.0$0.0231 of 4
shogi-014d3c+21.3100.0$0.03
shogi-02N*7e100.0100.0$0.02
shogi-035f5e100.0100.0$0.01
Scrabble14.133.3$0.0632 of 4
scrabble-0113B JOY42.2100.0$0.05
scrabble-02A10 UTE (illegal)0.00.0$0.05
scrabble-03D4 CEDAR (illegal)0.00.0$0.07
Kalah66.7100.0$0.0431 of 4
kalah-016100.0100.0<$0.01
kalah-0260.0100.0$0.03
kalah-032100.0100.0$0.08
Gomoku24.7100.0$0.0232 of 4
gomoku-01g937.5100.0$0.01
gomoku-02f31.9100.0$0.02
gomoku-03h1134.6100.0$0.03
Othello51.9100.0$0.0332 of 4
othello-01b1100.0100.0$0.04
othello-02a752.0100.0$0.03
othello-03h33.6100.0$0.03
Connect Four66.7100.0$0.0431 of 4
connect4-014100.0100.0$0.02
connect4-0260.0100.0$0.06
connect4-031100.0100.0$0.03
Quality is the mean move score × 100. n is answers: puzzles × samples. Open a game for each puzzle and the move played. Protocol boardbench-0.2.