Benchmark Almanac

CAIS Text Capabilities Index

Gemini 4 Argon 66.5 leads since Sep 2026 Not in model cards

Averages model scores on the text tests that the Center for AI Safety runs for its AI Dashboard. On 2026-11-01 the index listed Humanity's Last Exam and TextQuests.

One of three dashboard indexes, next to the Vision Capabilities Index and the Risk Index. The dashboard data also hold ARC-AGI-2, SWE-Bench Pro and HLE-Diamond scores. 67 models had entries on 2026-11-01.

Made by Center for AI Safety Released Dec 2025 Composite indices All tasks public Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Research frontier Researchers, on open problems or tasks that take days
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Short answer A number, word or short string matched to a key.
centerforaisafety/simple-evals 23 stars Leaderboard dashboard.safe.ai
66
Models scored on its own leaderboard

Gemini 4 Argon leads with 66.5

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best.

66 models. Scroll sideways for the rest.

  1. 66.5 Gemini 4 Argon
  2. 63.9 GPT-6 Astra
  3. 61.4 GPT-6.1 Sol
  4. 56.3 Claude Opus 5.5
  5. 55.6 Claude Fable 5.1
  6. 54.4 Claude Fable 5
  7. 52.2 Claude Opus 5
  8. 50.2 Gemini 3.8 Flash
  9. 49.3 Gemini 3.7 Flash
  10. 48.8 GPT-5.6 Sol
  11. 47.6 Muse Wonder
  12. 46.4 Kimi K3.1 Pro
  13. 46.3 GPT-5.5
  14. 45.8 Gemini 3.1 Pro
  15. 43.5 Grok 4.6
  16. 41.3 GPT-5.4
  17. 41.2 Claude Opus 4.8
  18. 40.5 Kimi K3
  19. 39.7 Gemini 3 Pro
  20. 39.1 Gemini 3.5 Flash
  21. 38.2 GPT-5.6 Terra
  22. 38.1 Muse Spark 1.3
  23. 38.0 Claude Opus 4.7
  24. 37.5 Claude Opus 4.6
  25. 37.4 GLM-5.2
  26. 37.2 Muse Spark 1.1
  27. 36.5 Gemini 3 Flash
  28. 36.3 Muse Spark Open
  29. 35.6 GLM-5.3
  30. 33.5 GPT-5.2
  31. 32.3 Claude Opus 4.5
  32. 32.3 Grok 4.5
  33. 31.4 Kimi K2.6
  34. 31.3 GPT-5.6 Luna
  35. 30.7 GPT-5.1
  36. 27.7 GPT-5
  37. 27.6 GLM-5.1
  38. 26.6 GPT-5.4 Mini
  39. 26.4 DeepSeek-V4-Pro
  40. 26.3 Claude Sonnet 4.6
  41. 26.2 Grok 4
  42. 25.6 o3
  43. 24.9 Kimi K2.5
  44. 24.3 Grok 4.20
  45. 23.4 Grok 4.3
  46. 22.9 Claude Sonnet 4.5
  47. 22.4 Gemini 2.5 Pro
  48. 21.5 DeepSeek-V3.2
  49. 19.9 Kimi K2 Thinking
  50. 19.0 Grok 4 Fast
  51. 18.8 Gemini 2.5 Pro
  52. 18.5 GPT-5 mini
  53. 18.1 Gemini 3.1 Flash-Lite
  54. 17.4 Grok 4.1 Fast
  55. 17.1 Claude Sonnet 4
  56. 16.1 GPT-5.4 Nano
  57. 15.4 o4-mini
  58. 13.2 Gemini 2.5 Flash
  59. 12.7 o3-mini
  60. 12.4 Claude Haiku 4.5
  61. 11.8 DeepSeek-R1
  62. 9.20 Gemini 2.5 Flash-Lite
  63. 8.00 o1
  64. 8.00 Claude 3.7 Sonnet
  65. 7.90 GPT-4o
  66. 6.60 GPT-5 nano
  1. 1 Gemini 4 Argon high 66.5
  2. 2 GPT-6 Astra high 63.9
  3. 3 GPT-6.1 Sol high 61.4
  4. 4 Claude Opus 5.5 high 56.3
  5. 5 Claude Fable 5.1 high 55.6
  6. 6 Claude Fable 5 high 54.4
  7. 7 Claude Opus 5 high 52.2
  8. 8 Gemini 3.8 Flash high 50.2
  9. 9 Gemini 3.7 Flash high 49.3
  10. 10 GPT-5.6 Sol high 48.8
  11. 11 Muse Wonder high 47.6
  12. 12 Kimi K3.1 Pro high 46.4
  13. 13 GPT-5.5 high 46.3
  14. 14 Gemini 3.1 Pro high 45.8
  15. 15 Grok 4.6 43.5
  16. 16 GPT-5.4 high 41.3
  17. 17 Claude Opus 4.8 high 41.2
  18. 18 Kimi K3 high 40.5
  19. 19 Gemini 3 Pro high 39.7
  20. 20 Gemini 3.5 Flash 39.1
  21. 21 GPT-5.6 Terra high 38.2
  22. 22 Muse Spark 1.3 high 38.1
  23. 23 Claude Opus 4.7 38.0
  24. 24 Claude Opus 4.6 37.5
  25. 25 GLM-5.2 37.4
  26. 26 Muse Spark 1.1 high 37.2
  27. 27 Gemini 3 Flash high 36.5
  28. 28 Muse Spark Open high 36.3
  29. 29 GLM-5.3 35.6
  30. 30 GPT-5.2 high 33.5
  31. 31 Claude Opus 4.5 thinking 32.3
  32. 32 Grok 4.5 32.3
  33. 33 Kimi K2.6 31.4
  34. 34 GPT-5.6 Luna high 31.3
  35. 35 GPT-5.1 high 30.7
  36. 36 GPT-5 high 27.7
  37. 37 GLM-5.1 27.6
  38. 38 GPT-5.4 Mini high 26.6
  39. 39 DeepSeek-V4-Pro 26.4
  40. 40 Claude Sonnet 4.6 26.3
  41. 41 Grok 4 26.2
  42. 42 o3 high 25.6
  43. 43 Kimi K2.5 thinking 24.9
  44. 44 Grok 4.20 24.3
  45. 45 Grok 4.3 23.4
  46. 46 Claude Sonnet 4.5 thinking 22.9
  47. 47 Gemini 2.5 Pro Jun 2025, high 22.4
  48. 48 DeepSeek-V3.2 thinking 21.5
  49. 49 Kimi K2 Thinking thinking 19.9
  50. 50 Grok 4 Fast 19.0
  51. 51 Gemini 2.5 Pro Mar 2025, high 18.8
  52. 52 GPT-5 mini high 18.5
  53. 53 Gemini 3.1 Flash-Lite high 18.1
  54. 54 Grok 4.1 Fast 17.4
  55. 55 Claude Sonnet 4 thinking 17.1
  56. 56 GPT-5.4 Nano high 16.1
  57. 57 o4-mini high 15.4
  58. 58 Gemini 2.5 Flash Jun 2025, high 13.2
  59. 59 o3-mini high 12.7
  60. 60 Claude Haiku 4.5 thinking 12.4
  61. 61 DeepSeek-R1 11.8
  62. 62 Gemini 2.5 Flash-Lite Jun 2025, high 9.20
  63. 63 o1 8.00
  64. 64 Claude 3.7 Sonnet thinking 8.00
  65. 65 GPT-4o Nov 2024 7.90
  66. 66 GPT-5 nano high 6.60
#ModelScoreSettingDateSource
1 Gemini 4 Argon 66.5 high Sep 2026 Official leaderboard
2 GPT-6 Astra 63.9 high Sep 2026 Official leaderboard
3 GPT-6.1 Sol 61.4 high Sep 2026 Official leaderboard
4 Claude Opus 5.5 56.3 high Sep 2026 Official leaderboard
5 Claude Fable 5.1 55.6 high Sep 2026 Official leaderboard
6 Claude Fable 5 54.4 high Jun 2026 Official leaderboard
7 Claude Opus 5 52.2 high Jul 2026 Official leaderboard
8 Gemini 3.8 Flash 50.2 high Sep 2026 Official leaderboard
9 Gemini 3.7 Flash 49.3 high Aug 2026 Official leaderboard
10 GPT-5.6 Sol 48.8 high Jul 2026 Official leaderboard
11 Muse Wonder 47.6 high Oct 2026 Official leaderboard
12 Kimi K3.1 Pro 46.4 high Oct 2026 Official leaderboard
13 GPT-5.5 46.3 high Apr 2026 Official leaderboard
14 Gemini 3.1 Pro 45.8 high Feb 2026 Official leaderboard
15 Grok 4.6 43.5 Aug 2026 Official leaderboard
16 GPT-5.4 41.3 high Mar 2026 Official leaderboard
17 Claude Opus 4.8 41.2 high May 2026 Official leaderboard
18 Kimi K3 40.5 high Jul 2026 Official leaderboard
19 Gemini 3 Pro 39.7 high Nov 2025 Official leaderboard
20 Gemini 3.5 Flash 39.1 May 2026 Official leaderboard
21 GPT-5.6 Terra 38.2 high Jul 2026 Official leaderboard
22 Muse Spark 1.3 38.1 high Sep 2026 Official leaderboard
23 Claude Opus 4.7 38.0 Apr 2026 Official leaderboard
24 Claude Opus 4.6 37.5 Feb 2026 Official leaderboard
25 GLM-5.2 37.4 Jun 2026 Official leaderboard
26 Muse Spark 1.1 37.2 high Jul 2026 Official leaderboard
27 Gemini 3 Flash 36.5 high Dec 2025 Official leaderboard
28 Muse Spark Open 36.3 high Oct 2026 Official leaderboard
29 GLM-5.3 35.6 Aug 2026 Official leaderboard
30 GPT-5.2 33.5 high Dec 2025 Official leaderboard
31 Claude Opus 4.5 32.3 thinking Nov 2025 Official leaderboard
32 Grok 4.5 32.3 Jul 2026 Official leaderboard
33 Kimi K2.6 31.4 Apr 2026 Official leaderboard
34 GPT-5.6 Luna 31.3 high Jul 2026 Official leaderboard
35 GPT-5.1 30.7 high Nov 2025 Official leaderboard
36 GPT-5 27.7 high Aug 2025 Official leaderboard
37 GLM-5.1 27.6 Apr 2026 Official leaderboard
38 GPT-5.4 Mini 26.6 high Mar 2026 Official leaderboard
39 DeepSeek-V4-Pro 26.4 Apr 2026 Official leaderboard
40 Claude Sonnet 4.6 26.3 Feb 2026 Official leaderboard
41 Grok 4 26.2 Jul 2025 Official leaderboard
42 o3 25.6 high Apr 2025 Official leaderboard
43 Kimi K2.5 24.9 thinking Jan 2026 Official leaderboard
44 Grok 4.20 24.3 Feb 2026 Official leaderboard
45 Grok 4.3 23.4 Apr 2026 Official leaderboard
46 Claude Sonnet 4.5 22.9 thinking Sep 2025 Official leaderboard
47 Gemini 2.5 Pro Jun 2025 22.4 high Jun 2025 Official leaderboard
48 DeepSeek-V3.2 21.5 thinking Dec 2025 Official leaderboard
49 Kimi K2 Thinking 19.9 thinking Nov 2025 Official leaderboard
50 Grok 4 Fast 19.0 Sep 2025 Official leaderboard
51 Gemini 2.5 Pro Mar 2025 18.8 high Mar 2025 Official leaderboard
52 GPT-5 mini 18.5 high Aug 2025 Official leaderboard
53 Gemini 3.1 Flash-Lite 18.1 high Mar 2026 Official leaderboard
54 Grok 4.1 Fast 17.4 Nov 2025 Official leaderboard
55 Claude Sonnet 4 17.1 thinking May 2025 Official leaderboard
56 GPT-5.4 Nano 16.1 high Mar 2026 Official leaderboard
57 o4-mini 15.4 high Apr 2025 Official leaderboard
58 Gemini 2.5 Flash Jun 2025 13.2 high Jun 2025 Official leaderboard
59 o3-mini 12.7 high Jan 2025 Official leaderboard
60 Claude Haiku 4.5 12.4 thinking Oct 2025 Official leaderboard
61 DeepSeek-R1 11.8 Jan 2025 Official leaderboard
62 Gemini 2.5 Flash-Lite Jun 2025 9.20 high Jun 2025 Official leaderboard
63 o1 8.00 Dec 2024 Official leaderboard
64 Claude 3.7 Sonnet 8.00 thinking Feb 2025 Official leaderboard
65 GPT-4o Nov 2024 7.90 Nov 2024 Official leaderboard
66 GPT-5 nano 6.60 high Aug 2025 Official leaderboard

7 models have held number one since its release in Dec 2025

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

Google DeepMind OpenAI Anthropic
Own leaderboard 7 leaders
Gemini 4 Argon 66.5

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

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Other

All 41 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
  • 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
  • Qwen in-house English automatic evaluation Only in system cards Models answer Qwen in-house English datasets on knowledge, comprehension, coding, math and related skills. Skills and weights change between reports. Qwen2-72B 72.9%
  • Recognition of model-written user turns (Anthropic) Only in system cards Models write synthetic user turns for real Claude Code transcripts, then judge which version is real. Score is accuracy above chance. Claude Mythos Preview 0.35
  • LiveBench Dropped Models answer monthly refreshed questions across math, coding, reasoning, language, instruction following, and data analysis. Scored automatically against objective ground truth. Claude Fable 5.1 85.7%