SurveyBench: Analysis rating

Top 10 of 21 models · rating, higher is better · lines show the 95% range

  1. 1 Claude Fable 5.2Anthropic 1,664 (95% range 1,567–1,816)
  2. 2 Muse WonderMeta 1,471 (95% range 1,411–1,550)
  3. 3 Claude Opus 5.5Anthropic 1,447 (95% range 1,393–1,510)
  4. 3 GPT-6.1 SolOpenAI 1,447 (95% range 1,394–1,533)
  5. 5 Muse EmberMeta 1,392 (95% range 1,323–1,501)
  6. 6 Claude Haiku 5.5Anthropic 1,248 (95% range 1,206–1,307)
  7. 7 GPT-6.1 TerraOpenAI 1,247 (95% range 1,169–1,327)
  8. 8 DeepSeek-V4.1-Flash-1017DeepSeek 1,169 (95% range 1,098–1,241)
  9. 9 Particle-1Particle AI 1,154 (95% range 1,075–1,256)
  10. 10 GPT-6 LunaOpenAI 1,102 (95% range 1,009–1,192)

Results to 10 Nov 2026 · release 8342a7178929Source: Spring Prompt · CC BY 4.0