Benchmark Almanac

CAIS Risk Index

Claude Fable 5.2 8.40 leads since November 2026
Not in model cards

Averages seven safety measures that correlate weakly with general capability. They cover agent prompt injection, bioweapons help, political bias, overconfidence, lying under pressure and harmful or deceptive play in text games.

Members on November 1st, 2026: Agent Red Teaming, the VCT, HPCT and BioTIER refusal average, PCT, HLE miscalibration, MASK, MACHIAVELLI and TextQuests Harm. Lower is better.

Released
December 2025
Kind
Composite indices
Level
Research frontier
Scored
Checked automatically
Answers
Open-ended
Tasks
All public
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last model added
November 2026
Scores refreshed
November 12th, 2026
67
Models scored on its own leaderboard

Claude Fable 5.2 leads with 8.40

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.

  1. 1 Claude Fable 5.2 high 8.40
  2. 2 Claude Opus 5.5 high 26.2
  3. 3 o3 high 30.2
  4. 4 Claude Fable 5.1 high 30.3
  5. 5 Claude Fable 5 high 30.8
  6. 6 Muse Wonder high 32.4
  7. 7 Muse Spark 1.1 high 32.5
  8. 8 Gemini 4 Carbon high 33.4
  9. 9 GPT-6 Astra high 35.5
  10. 10 o4-mini high 35.9
  11. 11 Claude Opus 5 high 36.1
  12. 12 Claude Opus 4.7 37.8
  13. 13 Claude Opus 4.8 high 38.1
  14. 14 Muse Spark Open high 39.1
  15. 15 GPT-6.1 Sol high 39.9
  16. 16 Claude Sonnet 4.6 40.4
  17. 17 Claude Sonnet 4.5 thinking 40.5
  18. 18 Muse Spark 1.3 high 40.6
  19. 19 Claude Opus 4.5 thinking 41.6
  20. 20 GPT-5.5 high 42.4
  21. 21 Gemini 4 Argon high 42.6
  22. 22 Claude Opus 4.6 43.2
  23. 23 GPT-5.6 Sol high 44.5
  24. 24 GPT-5 high 44.9
  25. 25 Claude 3.7 Sonnet thinking 46.0
  26. 26 GPT-5.6 Terra high 46.7
  27. 27 Claude Sonnet 4 thinking 47.1
  28. 28 Grok 4.6 48.8
  29. 29 GLM-5.3 48.8
  30. 30 GPT-5.4 high 49.2
  31. 31 Gemini 3.8 Flash high 49.3
  32. 32 GPT-5.2 high 50.3
  33. 33 GPT-5.6 Luna high 50.3
  34. 34 Gemini 3.7 Flash high 51.1
  35. 35 Grok 4.20 51.7
  36. 36 GPT-5.1 high 51.8
  37. 37 GLM-5.2 51.9
  38. 38 Grok 4.5 53.4
  39. 39 Kimi K3 high 53.6
  40. 40 GPT-5.4 Mini high 53.9
  41. 41 Claude Haiku 4.5 thinking 54.3
  42. 42 Grok 4.3 54.5
  43. 43 GPT-5 nano high 55.6
  44. 44 Gemini 3.1 Pro high 56.5
  45. 45 GPT-5.4 Nano high 56.8
  46. 46 GLM-5.1 58.0
  47. 47 Grok 4 Fast 58.3
  48. 48 GPT-5 mini high 58.5
  49. 49 Gemini 2.5 Pro Mar 2025, high 59.0
  50. 50 Grok 4 59.3
  51. 51 Kimi K2.6 59.5
  52. 52 Grok 4.1 Fast 60.0
  53. 53 Gemini 3.1 Flash-Lite high 61.0
  54. 54 Gemini 3.5 Flash 61.1
  55. 55 Kimi K2 Thinking thinking 63.9
  56. 56 o1 64.0
  57. 57 Gemini 3 Pro high 64.0
  58. 58 Gemini 2.5 Pro Jun 2025, high 64.3
  59. 59 Kimi K2.5 thinking 65.5
  60. 60 DeepSeek-V4-Pro 65.5
  61. 61 Gemini 3 Flash high 66.3
  62. 62 DeepSeek-R1 66.9
  63. 63 o3-mini high 67.6
  64. 64 DeepSeek-V3.2 thinking 68.3
  65. 65 Gemini 2.5 Flash Jun 2025, high 70.0
  66. 66 GPT-4o Nov 2024 74.4
  67. 67 Gemini 2.5 Flash-Lite Jun 2025, high 77.1

Source: Official CAIS Risk Index leaderboard

#ModelScoreSettingDateSource
1 Claude Fable 5.2 8.40 high Nov 2026 Official leaderboard
2 Claude Opus 5.5 26.2 high Sep 2026 Official leaderboard
3 o3 30.2 high Apr 2025 Official leaderboard
4 Claude Fable 5.1 30.3 high Sep 2026 Official leaderboard
5 Claude Fable 5 30.8 high Jun 2026 Official leaderboard
6 Muse Wonder 32.4 high Oct 2026 Official leaderboard
7 Muse Spark 1.1 32.5 high Jul 2026 Official leaderboard
8 Gemini 4 Carbon 33.4 high Oct 2026 Official leaderboard
9 GPT-6 Astra 35.5 high Sep 2026 Official leaderboard
10 o4-mini 35.9 high Apr 2025 Official leaderboard
11 Claude Opus 5 36.1 high Jul 2026 Official leaderboard
12 Claude Opus 4.7 37.8 Apr 2026 Official leaderboard
13 Claude Opus 4.8 38.1 high May 2026 Official leaderboard
14 Muse Spark Open 39.1 high Oct 2026 Official leaderboard
15 GPT-6.1 Sol 39.9 high Sep 2026 Official leaderboard
16 Claude Sonnet 4.6 40.4 Feb 2026 Official leaderboard
17 Claude Sonnet 4.5 40.5 thinking Sep 2025 Official leaderboard
18 Muse Spark 1.3 40.6 high Sep 2026 Official leaderboard
19 Claude Opus 4.5 41.6 thinking Nov 2025 Official leaderboard
20 GPT-5.5 42.4 high Apr 2026 Official leaderboard
21 Gemini 4 Argon 42.6 high Sep 2026 Official leaderboard
22 Claude Opus 4.6 43.2 Feb 2026 Official leaderboard
23 GPT-5.6 Sol 44.5 high Jul 2026 Official leaderboard
24 GPT-5 44.9 high Aug 2025 Official leaderboard
25 Claude 3.7 Sonnet 46.0 thinking Feb 2025 Official leaderboard
26 GPT-5.6 Terra 46.7 high Jul 2026 Official leaderboard
27 Claude Sonnet 4 47.1 thinking May 2025 Official leaderboard
28 Grok 4.6 48.8 Aug 2026 Official leaderboard
29 GLM-5.3 48.8 Aug 2026 Official leaderboard
30 GPT-5.4 49.2 high Mar 2026 Official leaderboard
31 Gemini 3.8 Flash 49.3 high Sep 2026 Official leaderboard
32 GPT-5.2 50.3 high Dec 2025 Official leaderboard
33 GPT-5.6 Luna 50.3 high Jul 2026 Official leaderboard
34 Gemini 3.7 Flash 51.1 high Aug 2026 Official leaderboard
35 Grok 4.20 51.7 Feb 2026 Official leaderboard
36 GPT-5.1 51.8 high Nov 2025 Official leaderboard
37 GLM-5.2 51.9 Jun 2026 Official leaderboard
38 Grok 4.5 53.4 Jul 2026 Official leaderboard
39 Kimi K3 53.6 high Jul 2026 Official leaderboard
40 GPT-5.4 Mini 53.9 high Mar 2026 Official leaderboard
41 Claude Haiku 4.5 54.3 thinking Oct 2025 Official leaderboard
42 Grok 4.3 54.5 Apr 2026 Official leaderboard
43 GPT-5 nano 55.6 high Aug 2025 Official leaderboard
44 Gemini 3.1 Pro 56.5 high Feb 2026 Official leaderboard
45 GPT-5.4 Nano 56.8 high Mar 2026 Official leaderboard
46 GLM-5.1 58.0 Apr 2026 Official leaderboard
47 Grok 4 Fast 58.3 Sep 2025 Official leaderboard
48 GPT-5 mini 58.5 high Aug 2025 Official leaderboard
49 Gemini 2.5 Pro Mar 2025 59.0 high Mar 2025 Official leaderboard
50 Grok 4 59.3 Jul 2025 Official leaderboard
51 Kimi K2.6 59.5 Apr 2026 Official leaderboard
52 Grok 4.1 Fast 60.0 Nov 2025 Official leaderboard
53 Gemini 3.1 Flash-Lite 61.0 high Mar 2026 Official leaderboard
54 Gemini 3.5 Flash 61.1 May 2026 Official leaderboard
55 Kimi K2 Thinking 63.9 thinking Nov 2025 Official leaderboard
56 o1 64.0 Dec 2024 Official leaderboard
57 Gemini 3 Pro 64.0 high Nov 2025 Official leaderboard
58 Gemini 2.5 Pro Jun 2025 64.3 high Jun 2025 Official leaderboard
59 Kimi K2.5 65.5 thinking Jan 2026 Official leaderboard
60 DeepSeek-V4-Pro 65.5 Apr 2026 Official leaderboard
61 Gemini 3 Flash 66.3 high Dec 2025 Official leaderboard
62 DeepSeek-R1 66.9 Jan 2025 Official leaderboard
63 o3-mini 67.6 high Jan 2025 Official leaderboard
64 DeepSeek-V3.2 68.3 thinking Dec 2025 Official leaderboard
65 Gemini 2.5 Flash Jun 2025 70.0 high Jun 2025 Official leaderboard
66 GPT-4o Nov 2024 74.4 Nov 2024 Official leaderboard
67 Gemini 2.5 Flash-Lite Jun 2025 77.1 high Jun 2025 Official leaderboard

3 models have held number one on the official leaderboard since its release in December 2025

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI Anthropic
Own leaderboard 3 leaders
Claude Fable 5.2 8.40

Source: Official CAIS Risk Index leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
0.00 20.0 40.0 60.0 80.0 2025 2026 2027 0.00 20.0 40.0 60.0 80.0 2025 2026 2027

Source: Official CAIS Risk Index leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Other

All 43 in Other
  • Personal Superintelligence Index (Meta) In use Meta's composite of seven personal and knowledge-work evals: Agents' Last Exam, AutomationBench-AA, FocusBench, JobBench, GDP.pdf, VBT and DRACO.
  • Artificial Analysis Intelligence Index v5 In use Version 5 of Artificial Analysis's index: nine evaluations in four equal categories, most with private test sets, run three times each in its own agent harness. Gemini 4 Carbon 59.0
  • SAGE v1.1 (Vals) In use Models grade handwritten undergraduate math and computer science work against rubrics. Scored by class-balanced accuracy of each rubric check against expert grading. Claude Opus 5.5 91.3%
  • MTEB Fading Massive Text Embedding Benchmark: retrieval, clustering, classification across languages and modalities. sonar 42%
  • Anthropic ECI (AECI) Only in system cards Index that combines a basket of capability benchmarks into one capability score to track trends across Anthropic models. Claude Opus 5.5 169.4
  • SPEED-Bench Only in system cards Measures speculative decoding across varied prompts and domains. Reports the average acceptance length of draft tokens, so higher means faster generation at the same output. Nemotron 3 Ultra 5.15