DeckBench: Deck rating

Top 10 of 36 models · rating, higher is better · lines show the 95% range · ≈ cannot be told apart from the leader

  1. 1≈ Qwen4-MaxAlibaba 1,563 (95% range 1,478–1,692)
  2. 2≈ GPT-6 AstraOpenAI 1,551 (95% range 1,434–1,730)
  3. 3≈ GPT-6 SolOpenAI 1,438 (95% range 1,370–1,545)
  4. 4≈ Claude Fable 5.2Anthropic 1,403 (95% range 1,311–1,512)
  5. 5 GPT-6.1 SolOpenAI 1,375 (95% range 1,305–1,463)
  6. 6≈ GPT-6.2 SolOpenAI 1,370 (95% range 1,277–1,493)
  7. 7 Claude Sonnet 5.5Anthropic 1,334 (95% range 1,249–1,445)
  8. 8 Claude Opus 5.5Anthropic 1,323 (95% range 1,245–1,406)
  9. 9 GPT-6.1 TerraOpenAI 1,280 (95% range 1,196–1,390)
  10. 10 MiniMax M3.1 FlashMiniMax 1,266 (95% range 1,177–1,367)

Results to 17 Nov 2026 · release 3608133e8331Source: Spring Prompt · CC BY 4.0