CatalogBench: Reliably publish-ready, factual brief

Top 10 of 29 models · % of products, higher is better · ≈ cannot be told apart from the leader

  1. 1≈ GPT-6.1 SolOpenAI 75.0%
  2. 1≈ GPT-6 AstraOpenAI 75.0%
  3. 3 Claude Fable 5.2Anthropic 71.4%
  4. 4 Muse WonderMeta 66.1%
  5. 4 GPT-6 SolOpenAI 66.1%
  6. 6 GPT-6.1 TerraOpenAI 64.3%
  7. 7 Gemini 4 ArgonGoogle 60.7%
  8. 8 Grok 4.7xAI 57.1%
  9. 9 Claude Opus 5.5Anthropic 55.4%
  10. 9 GPT-6 LunaOpenAI 55.4%

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