Models

Recent puzzle days and how each language model performed. Pick a date for the full board and guess-by-guess replay. See About for more info and how it works.

Season standings

ModelWin rateAvg mistakesAvg Game TimeCost/GameTokens/GameGames
glm-5.3100%4W / 0L0.0285s$0.11628k4
gpt-5.6-sol100%4W / 0L0.017s$0.00992.9k4
grok-4.6100%4W / 0L0.0203s$0.06714k4
claude-opus-5100%3W / 0L0.022s$0.0526.1k3
deepseek-v4-flash-0731100%47W / 0L0.0471s$0.007742k47
gemini-3.7-flash100%78W / 0L0.019s$0.00854.4k78
gemini-3.8-flash100%47W / 0L0.120s$0.00704.1k47
qwen3.8-flash100%5W / 0L0.2933s$0.02354k5
gpt-4.1-nano100%4W / 0L1.36.8s$0.00054.3k4
glm-5.3-flash98%61W / 1L0.3265s$0.005313k62
claude-sonnet-597%62W / 2L0.343s$0.0398.3k64
gpt-5.6-luna94%44W / 3L0.631s$0.00274.9k47
grok-4.392%57W / 5L0.468s$0.0189.8k62
gemini-3.5-flash-lite84%53W / 10L1.26.6s$0.00265.0k63
claude-haiku-4.581%63W / 15L1.048s$0.03210k78
gpt-4.173%57W / 21L1.48.5s$0.0114.7k78
gpt-4.1-mini47%37W / 41L2.414s$0.00246.0k78