BulletBench: Ladder Elo

Top 10 of 26 models · ladder Elo, higher is better · lines show the 95% range · ≈ cannot be told apart from the leader

  1. 1≈ Gemini 3.5 Flash-LiteGoogle 892 (95% range 641–1,081)
  2. 2≈ Gemini 3.1 Flash-Lite (minimal reasoning)Google 780 (95% range 589–940)
  3. 3≈ Gemini 3.5 Flash (minimal reasoning)Google 727 (95% range 567–890)
  4. 4≈ Mercury DecideInception 614 (95% range 470–766)
  5. 5≈ GPT-6.1 Sol Ultrafast (low reasoning)OpenAI 584 (95% range 327–806)
  6. 6≈ d1Liquid AI 572 (95% range 428–718)
  7. 7≈ Seed-2.0-Mini (minimal reasoning)ByteDance 571 (95% range 420–710)
  8. 7≈ Mistral Small 4 (no reasoning)Mistral AI 571 (95% range 420–699)
  9. 7≈ GPT-5.4 mini (no reasoning)OpenAI 571 (95% range 412–721)
  10. 7≈ Inkling Small (no reasoning)Thinking Machines 571 (95% range 428–698)

Results to 9 Oct 2026 · release 8342a7178929Source: Spring Prompt · CC BY 4.0