Benchmark Almanac

Agent Arena

Claude Fable 5.2 28.6% leads since November 2026
Claude Fable 5.2 28.6% leads since November 2026
In use

Arena ranks models as the orchestrator in real Agent Mode sessions on arena.ai. Scores use behavioral signals such as file downloads, disapproval events, retries and steerability.

Arena says the board measures signals from real sessions rather than preference votes alone. It went live on June 4th, 2026, and signal definitions changed on October 1st, 2026 and October 12th, 2026. Scores can fall below zero, as Mistral Medium 3.5 does at -14.5.

Released
June 2026
Kind
General assistants
Level
Everyday
Scored
Ranked by votes
Answers
Agent task
Tasks
All public
Measures
Preference
Language
English
In and out
Text → Text, Actions
Interaction
Agentic
Environment
APIs and tools
Refresh
Live
Run by
A host
In views
Agents, Kept fresh
Strongest model tested
Claude Fable 5.2 #2 on the capability index
Last in a launch post
November 2026
Last model added
November 2026
Scores refreshed
November 18th, 2026
59
Entries on its own leaderboard, 58 models we track
1
Launch post lists it, from 1 lab
1
System card reports it, from 1 lab

Claude Fable 5.2 leads with 28.6% Claude Fable 5.2 leads with 28.6%

Best score per entry from the benchmark's own leaderboard, read on November 18th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

Claude Fable 5.2 leads with 28.6%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.

  1. 1 Claude Fable 5.2 max 28.6%
  2. 2 Claude Opus 5.5 max 18.2%
  3. 3 Gemini 4 Carbon high 16.4%
  4. 4 Claude Sonnet 5.5 max 15.3%
  5. 5 Gemini 4 Hydrogen high 14.9%
  6. 6 Claude Fable 5.1 max 14.1%
  7. 7 GPT-6 Astra max 13.9%
  8. 8 GPT-6.1 Sol max 11.8%
  9. 9 GPT-6.1 Terra max 11.3%
  10. 10 Gemini 4 Argon high 10%
  11. 11 Muse Wonder Full max 8.00%
  12. 12 Qwen4-Max xhigh 7.46%
  13. 13 GPT-6 Sol max 6.25%
  14. 14 Claude Fable 5 high 6.22%
  15. 15 Claude Opus 5 max 6.18%
  16. 16 Tencent Hy4 5.67%
  17. 17 Kimi K3.1 Pro max 3.95%
  18. 18 GPT-5.6 Sol xhigh 3.72%
  19. 19 Inkling 2 Small 2.51%
  20. 20 Grok 4.7 xhigh 1.78%
  21. 21 Muse Ember max 1.73%
  22. 22 MiMo-V2.6 Pro 1.50%
  23. 23 Gemini 3.8 Flash high 1.44%
  24. 24 Muse Spark 1.3 max 1.20%
  25. 25 Claude Sonnet 5 high 1.14%
  26. 26 DeepSeek V4.1 Flash max 1.13%
  27. 27 Muse Wonder max 0.95%
  28. 28 GPT-6 Luna max 0.79%
  29. 29 GLM-5.3 max 0.03%
  30. 30 Kimi K3 max -0.24%
  31. 31 Tencent Hy4 preview -0.26%
  32. 32 Claude Haiku 5.5 high -0.37%
  33. 33 GLM-5.2 max -0.49%
  34. 34 MiMo-V2.6 Flash -0.64%
  35. 35 Qwen 3.8 Max -0.87%
  36. 36 Grok 4.6 xhigh -0.92%
  37. 37 GPT-5.6 Terra xhigh -2.72%
  38. 38 Gemini 3.7 Flash high -3.06%
  39. 39 MiniMax-M3.1-Flash -3.52%
  40. 40 GLM-5.3-Flash -3.70%
  41. 41 Step 5 Preview -4.65%
  42. 42 Qwen3.8-Flash-Next -5.18%
  43. 43 Qwen 3.8 27B -5.63%
  44. 44 Gemini 3.1 Pro -5.64%
  45. 45 Mistral Large 4 -6.81%
  46. 46 Beam 1.0 -7.40%
  47. 47 Qwen3.7-Max -8.01%
  48. 48 Tencent Hy3 -8.70%
  49. 49 Muse Spark 1.1 -9.01%
  50. 50 Qwen3.7-Plus -9.02%
  51. 51 Muse Spark 1.2 xhigh -9.08%
  52. 52 MiniMax-M3 -11.18%
  53. 53 Spark-X2.6 -12.25%
  54. 54 Mistral Medium 3.5 -14.47%
  55. 55 Nemotron 3 Ultra -16.40%
  56. 56 Nemotron 3.5 Super VL nvfp4 -16.73%
  57. 57 Inkling -18.28%
  58. 58 Inkling-Small -18.70%
  59. 59 Solar Pro 4 -23.93%

Source: Official Agent Arena leaderboard

  1. 1 Claude Fable 5.2 max 28.6%
  2. 2 Claude Opus 5.5 max 18.2%
  3. 3 Gemini 4 Carbon high 16.4%
  4. 4 Claude Sonnet 5.5 max 15.3%
  5. 5 Gemini 4 Hydrogen high 14.9%
  6. 6 Claude Fable 5.1 max 14.1%
  7. 7 GPT-6 Astra max 13.9%
  8. 8 GPT-6.1 Sol max 11.8%
  9. 9 GPT-6.1 Terra max 11.3%
  10. 10 Gemini 4 Argon high 10%
  11. 11 Muse Wonder Full max 8.00%
  12. 12 Qwen4-Max xhigh 7.46%
  13. 13 GPT-6 Sol max 6.25%
  14. 14 Claude Fable 5 high 6.22%
  15. 15 Claude Opus 5 max 6.18%
  16. 16 Tencent Hy4 5.67%
  17. 17 Kimi K3.1 Pro max 3.95%
  18. 18 GPT-5.6 Sol xhigh 3.72%
  19. 19 Inkling 2 Small 2.51%
  20. 20 Grok 4.7 xhigh 1.78%
  21. 21 Muse Ember max 1.73%
  22. 22 MiMo-V2.6 Pro 1.50%
  23. 23 Gemini 3.8 Flash high 1.44%
  24. 24 Muse Spark 1.3 max 1.20%
  25. 25 Claude Sonnet 5 high 1.14%
  26. 26 DeepSeek V4.1 Flash max 1.13%
  27. 27 Muse Wonder max 0.95%
  28. 28 GPT-6 Luna max 0.79%
  29. 29 GLM-5.3 max 0.03%
  30. 30 Kimi K3 max -0.24%
  31. 31 Tencent Hy4 preview -0.26%
  32. 32 Claude Haiku 5.5 high -0.37%
  33. 33 GLM-5.2 max -0.49%
  34. 34 MiMo-V2.6 Flash -0.64%
  35. 35 Qwen 3.8 Max -0.87%
  36. 36 Grok 4.6 xhigh -0.92%
  37. 37 GPT-5.6 Terra xhigh -2.72%
  38. 38 Gemini 3.7 Flash high -3.06%
  39. 39 MiniMax-M3.1-Flash -3.52%
  40. 40 GLM-5.3-Flash -3.70%
  41. 41 Step 5 Preview -4.65%
  42. 42 Qwen3.8-Flash-Next -5.18%
  43. 43 Qwen 3.8 27B -5.63%
  44. 44 Gemini 3.1 Pro -5.64%
  45. 45 Mistral Large 4 -6.81%
  46. 46 Beam 1.0 -7.40%
  47. 47 Qwen3.7-Max -8.01%
  48. 48 Tencent Hy3 -8.70%
  49. 49 Muse Spark 1.1 -9.01%
  50. 50 Qwen3.7-Plus -9.02%
  51. 51 Muse Spark 1.2 xhigh -9.08%
  52. 52 MiniMax-M3 -11.18%
  53. 53 Mistral Medium 3.5 -14.47%
  54. 54 Nemotron 3 Ultra -16.40%
  55. 55 Nemotron 3.5 Super VL nvfp4 -16.73%
  56. 56 Inkling -18.28%
  57. 57 Inkling-Small -18.70%
  58. 58 Solar Pro 4 -23.93%

Source: Official Agent Arena leaderboard

  1. 1 Claude Fable 5.2 max 28.6%
  2. 2 Claude Opus 5.5 max 18.2%
  3. 3 Gemini 4 Carbon high 16.4%
  4. 4 Claude Sonnet 5.5 max 15.3%
  5. 5 Gemini 4 Hydrogen high 14.9%
  6. 6 Claude Fable 5.1 max 14.1%
  7. 7 GPT-6 Astra max 13.9%
  8. 8 GPT-6.1 Sol max 11.8%
  9. 9 GPT-6.1 Terra max 11.3%
  10. 10 Gemini 4 Argon high 10%
  11. 11 Muse Wonder Full max 8.00%
  12. 12 Qwen4-Max 7.50%
  13. 13 GPT-6 Sol max 6.25%
  14. 14 Claude Fable 5 high 6.22%
  15. 15 Claude Opus 5 max 6.18%
  16. 16 Tencent Hy4 5.67%
  17. 17 Kimi K3.1 Pro max 3.95%
  18. 18 GPT-5.6 Sol xhigh 3.72%
  19. 19 Inkling 2 Small 2.51%
  20. 20 Grok 4.7 xhigh 1.78%
  21. 21 Muse Ember max 1.73%
  22. 22 MiMo-V2.6 Pro 1.50%
  23. 23 Gemini 3.8 Flash high 1.44%
  24. 24 Muse Spark 1.3 max 1.20%
  25. 25 Claude Sonnet 5 high 1.14%
  26. 26 DeepSeek V4.1 Flash max 1.13%
  27. 27 Muse Wonder max 0.95%
  28. 28 GPT-6 Luna max 0.79%
  29. 29 GLM-5.3 max 0.03%
  30. 30 Kimi K3 max -0.24%
  31. 31 Tencent Hy4 preview -0.26%
  32. 32 Claude Haiku 5.5 high -0.37%
  33. 33 GLM-5.2 max -0.49%
  34. 34 MiMo-V2.6 Flash -0.64%
  35. 35 Qwen 3.8 Max -0.87%
  36. 36 Grok 4.6 xhigh -0.92%
  37. 37 GPT-5.6 Terra xhigh -2.72%
  38. 38 Gemini 3.7 Flash high -3.06%
  39. 39 MiniMax-M3.1-Flash -3.52%
  40. 40 GLM-5.3-Flash -3.70%
  41. 41 Step 5 Preview -4.65%
  42. 42 Qwen3.8-Flash-Next -5.18%
  43. 43 Qwen 3.8 27B -5.63%
  44. 44 Gemini 3.1 Pro -5.64%
  45. 45 Mistral Large 4 -6.81%
  46. 46 Beam 1.0 -7.40%
  47. 47 Qwen3.7-Max -8.01%
  48. 48 Tencent Hy3 -8.70%
  49. 49 Muse Spark 1.1 -9.01%
  50. 50 Qwen3.7-Plus -9.02%
  51. 51 Muse Spark 1.2 xhigh -9.08%
  52. 52 MiniMax-M3 -11.18%
  53. 53 Mistral Medium 3.5 -14.47%
  54. 54 Nemotron 3 Ultra -16.40%
  55. 55 Nemotron 3.5 Super VL nvfp4 -16.73%
  56. 56 Inkling -18.28%
  57. 57 Inkling-Small -18.70%
  58. 58 Solar Pro 4 -23.93%

Sources: Official Agent Arena leaderboard and labs’ model cards

#EntryScore ± 95% CISettingReleasedRun onSource
1 Claude Fable 5.2 28.6% ±3.7 max Nov 2026 Nov 2026 Official leaderboard
2 Claude Opus 5.5 18.2% ±2 max Sep 2026 Nov 2026 Official leaderboard
3 Gemini 4 Carbon 16.4% ±0.7 high Oct 2026 Nov 2026 Official leaderboard
4 Claude Sonnet 5.5 15.3% ±2.9 max Sep 2026 Nov 2026 Official leaderboard
5 Gemini 4 Hydrogen 14.9% ±1.9 high Nov 2026 Nov 2026 Official leaderboard
6 Claude Fable 5.1 14.1% ±2.1 max Sep 2026 Nov 2026 Official leaderboard
7 GPT-6 Astra 13.9% ±2 max Sep 2026 Nov 2026 Official leaderboard
8 GPT-6.1 Sol 11.8% ±1.1 max Sep 2026 Nov 2026 Official leaderboard
9 GPT-6.1 Terra 11.3% ±1.2 max Oct 2026 Nov 2026 Official leaderboard
10 Gemini 4 Argon 10% ±0.9 high Sep 2026 Nov 2026 Official leaderboard
11 Muse Wonder Full 8.00% ±0.9 max Oct 2026 Nov 2026 Official leaderboard
12 Qwen4-Max 7.46% ±1.3 xhigh Nov 2026 Nov 2026 Official leaderboard
13 GPT-6 Sol 6.25% ±1.6 max Sep 2026 Nov 2026 Official leaderboard
14 Claude Fable 5 6.22% ±1.2 high Jun 2026 Nov 2026 Official leaderboard
15 Claude Opus 5 6.18% ±1.3 max Jul 2026 Nov 2026 Official leaderboard
16 Tencent Hy4 5.67% ±1.3 Nov 2026 Nov 2026 Official leaderboard
17 Kimi K3.1 Pro 3.95% ±1.1 max Oct 2026 Nov 2026 Official leaderboard
18 GPT-5.6 Sol 3.72% ±1.1 xhigh Jul 2026 Nov 2026 Official leaderboard
19 Inkling 2 Small 2.51% ±1.3 Nov 2026 Nov 2026 Official leaderboard
20 Grok 4.7 1.78% ±1.5 xhigh Sep 2026 Nov 2026 Official leaderboard
21 Muse Ember 1.73% ±1.5 max Nov 2026 Nov 2026 Official leaderboard
22 MiMo-V2.6 Pro 1.50% ±0.9 Sep 2026 Nov 2026 Official leaderboard
23 Gemini 3.8 Flash 1.44% ±0.9 high Sep 2026 Nov 2026 Official leaderboard
24 Muse Spark 1.3 1.20% ±0.9 max Sep 2026 Nov 2026 Official leaderboard
25 Claude Sonnet 5 1.14% ±1.2 high Jun 2026 Nov 2026 Official leaderboard
26 DeepSeek V4.1 Flash 1.13% ±0.5 max Sep 2026 Nov 2026 Official leaderboard
27 Muse Wonder 0.95% ±1.3 max Oct 2026 Nov 2026 Official leaderboard
28 GPT-6 Luna 0.79% ±0.5 max Sep 2026 Nov 2026 Official leaderboard
29 GLM-5.3 0.03% ±0.8 max Aug 2026 Nov 2026 Official leaderboard
30 Kimi K3 -0.24% ±0.6 max Jul 2026 Nov 2026 Official leaderboard
31 Tencent Hy4 preview -0.26% ±0.7 Aug 2026 Nov 2026 Official leaderboard
32 Claude Haiku 5.5 -0.37% ±1 high Oct 2026 Nov 2026 Official leaderboard
33 GLM-5.2 -0.49% ±0.8 max Jun 2026 Nov 2026 Official leaderboard
34 MiMo-V2.6 Flash -0.64% ±1 Sep 2026 Nov 2026 Official leaderboard
35 Qwen 3.8 Max -0.87% ±0.7 Aug 2026 Nov 2026 Official leaderboard
36 Grok 4.6 -0.92% ±1.3 xhigh Aug 2026 Nov 2026 Official leaderboard
37 GPT-5.6 Terra -2.72% ±0.8 xhigh Jul 2026 Nov 2026 Official leaderboard
38 Gemini 3.7 Flash -3.06% ±0.9 high Aug 2026 Nov 2026 Official leaderboard
39 MiniMax-M3.1-Flash -3.52% ±1.5 Oct 2026 Nov 2026 Official leaderboard
40 GLM-5.3-Flash -3.70% ±0.7 Aug 2026 Nov 2026 Official leaderboard
41 Step 5 Preview -4.65% ±1.4 Sep 2026 Nov 2026 Official leaderboard
42 Qwen3.8-Flash-Next -5.18% ±0.8 Aug 2026 Nov 2026 Official leaderboard
43 Qwen 3.8 27B -5.63% ±0.9 Aug 2026 Nov 2026 Official leaderboard
44 Gemini 3.1 Pro -5.64% ±1.7 Feb 2026 Nov 2026 Official leaderboard
45 Mistral Large 4 -6.81% ±2 Oct 2026 Nov 2026 Official leaderboard
46 Beam 1.0 -7.40% ±2.9 Oct 2026 Nov 2026 Official leaderboard
47 Qwen3.7-Max -8.01% ±1.3 May 2026 Nov 2026 Official leaderboard
48 Tencent Hy3 -8.70% ±1.5 Jul 2026 Nov 2026 Official leaderboard
49 Muse Spark 1.1 -9.01% ±0.7 Jul 2026 Nov 2026 Official leaderboard
50 Qwen3.7-Plus -9.02% ±1.6 Jun 2026 Nov 2026 Official leaderboard
51 Muse Spark 1.2 -9.08% ±0.9 xhigh Aug 2026 Nov 2026 Official leaderboard
52 MiniMax-M3 -11.18% ±1.2 Jun 2026 Nov 2026 Official leaderboard
53 Spark-X2.6 -12.25% ±1.3 Nov 2026 Official leaderboard
54 Mistral Medium 3.5 -14.47% ±1.8 Apr 2026 Nov 2026 Official leaderboard
55 Nemotron 3 Ultra -16.40% ±3.2 Jun 2026 Nov 2026 Official leaderboard
56 Nemotron 3.5 Super VL -16.73% ±0.7 nvfp4 Oct 2026 Nov 2026 Official leaderboard
57 Inkling -18.28% ±1.6 Jul 2026 Nov 2026 Official leaderboard
58 Inkling-Small -18.70% ±2 Jul 2026 Nov 2026 Official leaderboard
59 Solar Pro 4 -23.93% ±2.2 Aug 2026 Nov 2026 Official leaderboard
#ModelScore ± 95% CISettingReleasedRun onSource
1 Claude Fable 5.2 28.6% ±3.7 max Nov 2026 Nov 2026 Official leaderboard
2 Claude Opus 5.5 18.2% ±2 max Sep 2026 Nov 2026 Official leaderboard
3 Gemini 4 Carbon 16.4% ±0.7 high Oct 2026 Nov 2026 Official leaderboard
4 Claude Sonnet 5.5 15.3% ±2.9 max Sep 2026 Nov 2026 Official leaderboard
5 Gemini 4 Hydrogen 14.9% ±1.9 high Nov 2026 Nov 2026 Official leaderboard
6 Claude Fable 5.1 14.1% ±2.1 max Sep 2026 Nov 2026 Official leaderboard
7 GPT-6 Astra 13.9% ±2 max Sep 2026 Nov 2026 Official leaderboard
8 GPT-6.1 Sol 11.8% ±1.1 max Sep 2026 Nov 2026 Official leaderboard
9 GPT-6.1 Terra 11.3% ±1.2 max Oct 2026 Nov 2026 Official leaderboard
10 Gemini 4 Argon 10% ±0.9 high Sep 2026 Nov 2026 Official leaderboard
11 Muse Wonder Full 8.00% ±0.9 max Oct 2026 Nov 2026 Official leaderboard
12 Qwen4-Max 7.46% ±1.3 xhigh Nov 2026 Nov 2026 Official leaderboard
13 GPT-6 Sol 6.25% ±1.6 max Sep 2026 Nov 2026 Official leaderboard
14 Claude Fable 5 6.22% ±1.2 high Jun 2026 Nov 2026 Official leaderboard
15 Claude Opus 5 6.18% ±1.3 max Jul 2026 Nov 2026 Official leaderboard
16 Tencent Hy4 5.67% ±1.3 Nov 2026 Nov 2026 Official leaderboard
17 Kimi K3.1 Pro 3.95% ±1.1 max Oct 2026 Nov 2026 Official leaderboard
18 GPT-5.6 Sol 3.72% ±1.1 xhigh Jul 2026 Nov 2026 Official leaderboard
19 Inkling 2 Small 2.51% ±1.3 Nov 2026 Nov 2026 Official leaderboard
20 Grok 4.7 1.78% ±1.5 xhigh Sep 2026 Nov 2026 Official leaderboard
21 Muse Ember 1.73% ±1.5 max Nov 2026 Nov 2026 Official leaderboard
22 MiMo-V2.6 Pro 1.50% ±0.9 Sep 2026 Nov 2026 Official leaderboard
23 Gemini 3.8 Flash 1.44% ±0.9 high Sep 2026 Nov 2026 Official leaderboard
24 Muse Spark 1.3 1.20% ±0.9 max Sep 2026 Nov 2026 Official leaderboard
25 Claude Sonnet 5 1.14% ±1.2 high Jun 2026 Nov 2026 Official leaderboard
26 DeepSeek V4.1 Flash 1.13% ±0.5 max Sep 2026 Nov 2026 Official leaderboard
27 Muse Wonder 0.95% ±1.3 max Oct 2026 Nov 2026 Official leaderboard
28 GPT-6 Luna 0.79% ±0.5 max Sep 2026 Nov 2026 Official leaderboard
29 GLM-5.3 0.03% ±0.8 max Aug 2026 Nov 2026 Official leaderboard
30 Kimi K3 -0.24% ±0.6 max Jul 2026 Nov 2026 Official leaderboard
31 Tencent Hy4 preview -0.26% ±0.7 Aug 2026 Nov 2026 Official leaderboard
32 Claude Haiku 5.5 -0.37% ±1 high Oct 2026 Nov 2026 Official leaderboard
33 GLM-5.2 -0.49% ±0.8 max Jun 2026 Nov 2026 Official leaderboard
34 MiMo-V2.6 Flash -0.64% ±1 Sep 2026 Nov 2026 Official leaderboard
35 Qwen 3.8 Max -0.87% ±0.7 Aug 2026 Nov 2026 Official leaderboard
36 Grok 4.6 -0.92% ±1.3 xhigh Aug 2026 Nov 2026 Official leaderboard
37 GPT-5.6 Terra -2.72% ±0.8 xhigh Jul 2026 Nov 2026 Official leaderboard
38 Gemini 3.7 Flash -3.06% ±0.9 high Aug 2026 Nov 2026 Official leaderboard
39 MiniMax-M3.1-Flash -3.52% ±1.5 Oct 2026 Nov 2026 Official leaderboard
40 GLM-5.3-Flash -3.70% ±0.7 Aug 2026 Nov 2026 Official leaderboard
41 Step 5 Preview -4.65% ±1.4 Sep 2026 Nov 2026 Official leaderboard
42 Qwen3.8-Flash-Next -5.18% ±0.8 Aug 2026 Nov 2026 Official leaderboard
43 Qwen 3.8 27B -5.63% ±0.9 Aug 2026 Nov 2026 Official leaderboard
44 Gemini 3.1 Pro -5.64% ±1.7 Feb 2026 Nov 2026 Official leaderboard
45 Mistral Large 4 -6.81% ±2 Oct 2026 Nov 2026 Official leaderboard
46 Beam 1.0 -7.40% ±2.9 Oct 2026 Nov 2026 Official leaderboard
47 Qwen3.7-Max -8.01% ±1.3 May 2026 Nov 2026 Official leaderboard
48 Tencent Hy3 -8.70% ±1.5 Jul 2026 Nov 2026 Official leaderboard
49 Muse Spark 1.1 -9.01% ±0.7 Jul 2026 Nov 2026 Official leaderboard
50 Qwen3.7-Plus -9.02% ±1.6 Jun 2026 Nov 2026 Official leaderboard
51 Muse Spark 1.2 -9.08% ±0.9 xhigh Aug 2026 Nov 2026 Official leaderboard
52 MiniMax-M3 -11.18% ±1.2 Jun 2026 Nov 2026 Official leaderboard
53 Mistral Medium 3.5 -14.47% ±1.8 Apr 2026 Nov 2026 Official leaderboard
54 Nemotron 3 Ultra -16.40% ±3.2 Jun 2026 Nov 2026 Official leaderboard
55 Nemotron 3.5 Super VL -16.73% ±0.7 nvfp4 Oct 2026 Nov 2026 Official leaderboard
56 Inkling -18.28% ±1.6 Jul 2026 Nov 2026 Official leaderboard
57 Inkling-Small -18.70% ±2 Jul 2026 Nov 2026 Official leaderboard
58 Solar Pro 4 -23.93% ±2.2 Aug 2026 Nov 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Fable 5.2 28.6% max Nov 2026 Official leaderboard
2 Claude Opus 5.5 18.2% max Sep 2026 Official leaderboard
3 Gemini 4 Carbon 16.4% high Oct 2026 Official leaderboard
4 Claude Sonnet 5.5 15.3% max Sep 2026 Official leaderboard
5 Gemini 4 Hydrogen 14.9% high Nov 2026 Official leaderboard
6 Claude Fable 5.1 14.1% max Sep 2026 Official leaderboard
7 GPT-6 Astra 13.9% max Sep 2026 Official leaderboard
8 GPT-6.1 Sol 11.8% max Sep 2026 Official leaderboard
9 GPT-6.1 Terra 11.3% max Oct 2026 Official leaderboard
10 Gemini 4 Argon 10% high Sep 2026 Official leaderboard
11 Muse Wonder Full 8.00% max Oct 2026 Official leaderboard
12 Qwen4-Max 7.50% Nov 2026 Lab model card
13 GPT-6 Sol 6.25% max Sep 2026 Official leaderboard
14 Claude Fable 5 6.22% high Jun 2026 Official leaderboard
15 Claude Opus 5 6.18% max Jul 2026 Official leaderboard
16 Tencent Hy4 5.67% Nov 2026 Official leaderboard
17 Kimi K3.1 Pro 3.95% max Oct 2026 Official leaderboard
18 GPT-5.6 Sol 3.72% xhigh Jul 2026 Official leaderboard
19 Inkling 2 Small 2.51% Nov 2026 Official leaderboard
20 Grok 4.7 1.78% xhigh Sep 2026 Official leaderboard
21 Muse Ember 1.73% max Nov 2026 Official leaderboard
22 MiMo-V2.6 Pro 1.50% Sep 2026 Official leaderboard
23 Gemini 3.8 Flash 1.44% high Sep 2026 Official leaderboard
24 Muse Spark 1.3 1.20% max Sep 2026 Official leaderboard
25 Claude Sonnet 5 1.14% high Jun 2026 Official leaderboard
26 DeepSeek V4.1 Flash 1.13% max Sep 2026 Official leaderboard
27 Muse Wonder 0.95% max Oct 2026 Official leaderboard
28 GPT-6 Luna 0.79% max Sep 2026 Official leaderboard
29 GLM-5.3 0.03% max Aug 2026 Official leaderboard
30 Kimi K3 -0.24% max Jul 2026 Official leaderboard
31 Tencent Hy4 preview -0.26% Aug 2026 Official leaderboard
32 Claude Haiku 5.5 -0.37% high Oct 2026 Official leaderboard
33 GLM-5.2 -0.49% max Jun 2026 Official leaderboard
34 MiMo-V2.6 Flash -0.64% Sep 2026 Official leaderboard
35 Qwen 3.8 Max -0.87% Aug 2026 Official leaderboard
36 Grok 4.6 -0.92% xhigh Aug 2026 Official leaderboard
37 GPT-5.6 Terra -2.72% xhigh Jul 2026 Official leaderboard
38 Gemini 3.7 Flash -3.06% high Aug 2026 Official leaderboard
39 MiniMax-M3.1-Flash -3.52% Oct 2026 Official leaderboard
40 GLM-5.3-Flash -3.70% Aug 2026 Official leaderboard
41 Step 5 Preview -4.65% Sep 2026 Official leaderboard
42 Qwen3.8-Flash-Next -5.18% Aug 2026 Official leaderboard
43 Qwen 3.8 27B -5.63% Aug 2026 Official leaderboard
44 Gemini 3.1 Pro -5.64% Feb 2026 Official leaderboard
45 Mistral Large 4 -6.81% Oct 2026 Official leaderboard
46 Beam 1.0 -7.40% Oct 2026 Official leaderboard
47 Qwen3.7-Max -8.01% May 2026 Official leaderboard
48 Tencent Hy3 -8.70% Jul 2026 Official leaderboard
49 Muse Spark 1.1 -9.01% Jul 2026 Official leaderboard
50 Qwen3.7-Plus -9.02% Jun 2026 Official leaderboard
51 Muse Spark 1.2 -9.08% xhigh Aug 2026 Official leaderboard
52 MiniMax-M3 -11.18% Jun 2026 Official leaderboard
53 Mistral Medium 3.5 -14.47% Apr 2026 Official leaderboard
54 Nemotron 3 Ultra -16.40% Jun 2026 Official leaderboard
55 Nemotron 3.5 Super VL -16.73% nvfp4 Oct 2026 Official leaderboard
56 Inkling -18.28% Jul 2026 Official leaderboard
57 Inkling-Small -18.70% Jul 2026 Official leaderboard
58 Solar Pro 4 -23.93% Aug 2026 Official leaderboard

Labs report scores for 1 model in their cards. Switch the scores above to rank them together.

Rankings by part

We read 5 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Bash reliability, Praise and complaint and Steerability. Claude Fable 5.2 leads 3 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Bash reliability

58 models we track have a score here.

  1. 1 Claude Sonnet 5 ↑24 14.4%
  2. 2 Claude Opus 5.5 14.3%
  3. 3 Claude Fable 5.2 ↓2 13.6%
  4. 4 GPT-6.1 Terra ↑5 12.8%
  5. 5 Claude Sonnet 5.5 ↓1 11.9%
  6. 6 GPT-5.6 Sol ↑12 10.5%
  7. 7 GPT-6 Sol ↑6 10.1%
  8. 8 Claude Fable 5 ↑6 9.57%
  9. 9 Claude Fable 5.1 ↓3 9.31%
  10. 10 Grok 4.6 ↑26 9.28%

Score against cost per task

Each cost is the median cost of one task in Agent Mode sessions, as Arena reports it. The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

-20% 0% 20% 40% $0.001$0.01$0.1$1$10 Median cost per task ($), log scale, lower is better Agent Arena score Claude Fable 5.2 (max)Claude Opus 5.5 (max)Gemini 4 Carbon (high)GPT-6.1 Sol (max)GPT-6.1 Terra (max)Gemini 4 Argon (high)GPT-6.1 Sol (medium)GPT-6.1 Terra (medium)Muse Ember (max)GPT-6 Luna (low) -20% 0% 20% 40% $0.001$0.01$0.1$1$10 Median cost per task ($), log scale, lower is better Agent Arena score Claude Fable 5.2 (max)Claude Opus 5.5 (max)Gemini 4 Carbon (high)GPT-6.1 Sol (max)GPT-6.1 Terra (max)Gemini 4 Argon (high)GPT-6.1 Sol (medium)GPT-6.1 Terra (medium)Muse Ember (max)GPT-6 Luna (low)
ModelSettingAgent Arena scoreMedian cost per task
Claude Fable 5.2 max 28.6% $9.02
Claude Opus 5.5 max 18.2% $5.65
Gemini 4 Carbon high 16.5% $1.35
Claude Sonnet 5.5 max 15.3% $3.46
Gemini 4 Hydrogen high 14.9% $1.53
Claude Fable 5.1 max 14.1% $7.83
GPT-6 Astra max 13.9% $5.28
Claude Opus 5.5 high 13.4% $2.12
GPT-6.1 Sol max 11.8% $0.99
GPT-6.1 Terra max 11.3% $0.87
Gemini 4 Argon high 10% $0.7
GPT-6.1 Sol medium 8% $0.37
Qwen4-Max xhigh 7.5% $1.47
Claude Sonnet 5.5 high 7.2% $0.88
GPT-6 Sol max 6.3% $1.85
Claude Fable 5 high 6.2% $3.33
Claude Opus 5 max 6.2% $4.5
Tencent Hy4 5.7% $0.57
GPT-6.1 Terra medium 4.4% $0.16
Kimi K3.1 Pro max 3.9% $1.16
GPT-5.6 Sol xhigh 3.7% $1.59
Inkling 2 Small 2.5% $0.26
Grok 4.7 xhigh 1.8% $3.2
Muse Ember max 1.7% $0.02
MiMo-V2.6 Pro 1.5% $0.16
Gemini 3.8 Flash high 1.4% $0.76
Muse Spark 1.3 max 1.2% $0.47
Claude Sonnet 5 high 1.1% $0.87
DeepSeek V4.1 Flash max 1.1% $0.14
Muse Wonder max 0.9% $1.16
GPT-6 Luna max 0.8% $0.08
GLM-5.3 max 0% $0.86
Kimi K3 max -0.2% $0.89
Tencent Hy4 preview -0.3% $0.43
Claude Haiku 5.5 high -0.4% $0.08
GLM-5.2 max -0.5% $0.86
MiMo-V2.6 Flash -0.6% $0.08
Qwen 3.8 Max -0.9% $1.24
Grok 4.6 xhigh -0.9% $1.33
GPT-5.6 Terra xhigh -2.7% $0.63
GPT-6 Luna medium -2.9% $0.02
Gemini 3.7 Flash high -3.1% $0.49
MiniMax-M3.1-Flash -3.5% $0.21
GLM-5.3-Flash -3.7% $0.14
Step 5 Preview -4.6% $0.7
Qwen3.8-Flash-Next -5.2% $0.23
Qwen 3.8 27B -5.6% $0.39
Gemini 3.1 Pro -5.6% $0.59
Mistral Large 4 -6.8% $0.32
GPT-6 Luna low -7.2% $0.01
Qwen3.7-Max -8% $0.28
Tencent Hy3 -8.7% $0.08
Qwen3.7-Plus -9% $0.17
Muse Spark 1.2 xhigh -9.1% $0.41
MiniMax-M3 -11.2% $0.38
Spark-X2.6 -12.2% $0.08
Mistral Medium 3.5 -14.5% $0.3
Nemotron 3 Ultra -16.4% $0.13
Inkling -18.3% $0.3
Inkling-Small -18.7% $0.16
Solar Pro 4 -23.9% $0.04

Score against cost per task

Each cost is the median cost of one task in Agent Mode sessions, as Arena reports it. The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

-20% 0% 20% 40% $0.001$0.01$0.1$1$10 Median cost per task ($), log scale, lower is better Agent Arena score Claude Fable 5.2 (max)Claude Opus 5.5 (max)Gemini 4 Carbon (high)GPT-6.1 Sol (max)GPT-6.1 Terra (max)Gemini 4 Argon (high)GPT-6.1 Sol (medium)GPT-6.1 Terra (medium)Muse Ember (max)GPT-6 Luna (low) -20% 0% 20% 40% $0.001$0.01$0.1$1$10 Median cost per task ($), log scale, lower is better Agent Arena score Claude Fable 5.2 (max)Claude Opus 5.5 (max)Gemini 4 Carbon (high)GPT-6.1 Sol (max)GPT-6.1 Terra (max)Gemini 4 Argon (high)GPT-6.1 Sol (medium)GPT-6.1 Terra (medium)Muse Ember (max)GPT-6 Luna (low)
ModelSettingAgent Arena scoreMedian cost per task
Claude Fable 5.2 max 28.6% $9.02
Claude Opus 5.5 max 18.2% $5.65
Gemini 4 Carbon high 16.5% $1.35
Claude Sonnet 5.5 max 15.3% $3.46
Gemini 4 Hydrogen high 14.9% $1.53
Claude Fable 5.1 max 14.1% $7.83
GPT-6 Astra max 13.9% $5.28
Claude Opus 5.5 high 13.4% $2.12
GPT-6.1 Sol max 11.8% $0.99
GPT-6.1 Terra max 11.3% $0.87
Gemini 4 Argon high 10% $0.7
GPT-6.1 Sol medium 8% $0.37
Qwen4-Max xhigh 7.5% $1.47
Claude Sonnet 5.5 high 7.2% $0.88
GPT-6 Sol max 6.3% $1.85
Claude Fable 5 high 6.2% $3.33
Claude Opus 5 max 6.2% $4.5
Tencent Hy4 5.7% $0.57
GPT-6.1 Terra medium 4.4% $0.16
Kimi K3.1 Pro max 3.9% $1.16
GPT-5.6 Sol xhigh 3.7% $1.59
Inkling 2 Small 2.5% $0.26
Grok 4.7 xhigh 1.8% $3.2
Muse Ember max 1.7% $0.02
MiMo-V2.6 Pro 1.5% $0.16
Gemini 3.8 Flash high 1.4% $0.76
Muse Spark 1.3 max 1.2% $0.47
Claude Sonnet 5 high 1.1% $0.87
DeepSeek V4.1 Flash max 1.1% $0.14
Muse Wonder max 0.9% $1.16
GPT-6 Luna max 0.8% $0.08
GLM-5.3 max 0% $0.86
Kimi K3 max -0.2% $0.89
Tencent Hy4 preview -0.3% $0.43
Claude Haiku 5.5 high -0.4% $0.08
GLM-5.2 max -0.5% $0.86
MiMo-V2.6 Flash -0.6% $0.08
Qwen 3.8 Max -0.9% $1.24
Grok 4.6 xhigh -0.9% $1.33
GPT-5.6 Terra xhigh -2.7% $0.63
GPT-6 Luna medium -2.9% $0.02
Gemini 3.7 Flash high -3.1% $0.49
MiniMax-M3.1-Flash -3.5% $0.21
GLM-5.3-Flash -3.7% $0.14
Step 5 Preview -4.6% $0.7
Qwen3.8-Flash-Next -5.2% $0.23
Qwen 3.8 27B -5.6% $0.39
Gemini 3.1 Pro -5.6% $0.59
Mistral Large 4 -6.8% $0.32
GPT-6 Luna low -7.2% $0.01
Qwen3.7-Max -8% $0.28
Tencent Hy3 -8.7% $0.08
Qwen3.7-Plus -9% $0.17
Muse Spark 1.2 xhigh -9.1% $0.41
MiniMax-M3 -11.2% $0.38
Mistral Medium 3.5 -14.5% $0.3
Nemotron 3 Ultra -16.4% $0.13
Inkling -18.3% $0.3
Inkling-Small -18.7% $0.16
Solar Pro 4 -23.9% $0.04

More effort adds a median 7 points

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Claude Opus 5.5
high max
18.2%
Claude Sonnet 5.5
high max
15.3%
GPT-6.1 Sol
medium max
11.8%
GPT-6.1 Terra
medium max
11.3%
GPT-6 Luna
low medium max
0.79%

Source: Official Agent Arena leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 max 28.6%
Claude Opus 5.5 max 18.2%
Gemini 4 Carbon high 16.4%
Claude Sonnet 5.5 max 15.3%
Gemini 4 Hydrogen high 14.9%
Claude Fable 5.1 max 14.1%
GPT-6 Astra max 13.9%
Claude Opus 5.5 high 13.3%
GPT-6.1 Sol max 11.8%
GPT-6.1 Terra max 11.3%
Gemini 4 Argon high 10%
GPT-6.1 Sol medium 8.04%
Muse Wonder Full max 8.00%
Qwen4-Max xhigh 7.46%
Claude Sonnet 5.5 high 7.18%
GPT-6 Sol max 6.25%
Claude Fable 5 high 6.22%
Claude Opus 5 max 6.18%
Tencent Hy4 5.67%
GPT-6.1 Terra medium 4.44%
Kimi K3.1 Pro max 3.95%
GPT-5.6 Sol xhigh 3.72%
Inkling 2 Small 2.51%
Grok 4.7 xhigh 1.78%
Muse Ember max 1.73%
MiMo-V2.6 Pro 1.50%
Gemini 3.8 Flash high 1.44%
Muse Spark 1.3 max 1.20%
Claude Sonnet 5 high 1.14%
DeepSeek V4.1 Flash max 1.13%
Muse Wonder max 0.95%
GPT-6 Luna max 0.79%
GLM-5.3 max 0.03%
Kimi K3 max -0.24%
Tencent Hy4 preview -0.26%
Claude Haiku 5.5 high -0.37%
GLM-5.2 max -0.49%
MiMo-V2.6 Flash -0.64%
Qwen 3.8 Max -0.87%
Grok 4.6 xhigh -0.92%
GPT-5.6 Terra xhigh -2.72%
GPT-6 Luna medium -2.89%
Gemini 3.7 Flash high -3.06%
MiniMax-M3.1-Flash -3.52%
GLM-5.3-Flash -3.70%
Step 5 Preview -4.65%
Qwen3.8-Flash-Next -5.18%
Qwen 3.8 27B -5.63%
Gemini 3.1 Pro -5.64%
Mistral Large 4 -6.81%
GPT-6 Luna low -7.18%
Beam 1.0 -7.40%
Qwen3.7-Max -8.01%
Tencent Hy3 -8.70%
Muse Spark 1.1 -9.01%
Qwen3.7-Plus -9.02%
Muse Spark 1.2 xhigh -9.08%
MiniMax-M3 -11.18%
Mistral Medium 3.5 -14.47%
Nemotron 3 Ultra -16.40%
Nemotron 3.5 Super VL nvfp4 -16.73%
Inkling -18.28%
Inkling-Small -18.70%
Solar Pro 4 -23.93%

More effort adds a median 7 points

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Claude Opus 5.5
high max
18.2%
Claude Sonnet 5.5
high max
15.3%
GPT-6.1 Sol
medium max
11.8%
GPT-6.1 Terra
medium max
11.3%
GPT-6 Luna
low medium max
0.79%

Source: Official Agent Arena leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 max 28.6%
Claude Opus 5.5 max 18.2%
Gemini 4 Carbon high 16.4%
Claude Sonnet 5.5 max 15.3%
Gemini 4 Hydrogen high 14.9%
Claude Fable 5.1 max 14.1%
GPT-6 Astra max 13.9%
Claude Opus 5.5 high 13.3%
GPT-6.1 Sol max 11.8%
GPT-6.1 Terra max 11.3%
Gemini 4 Argon high 10%
GPT-6.1 Sol medium 8.04%
Muse Wonder Full max 8.00%
Qwen4-Max xhigh 7.46%
Claude Sonnet 5.5 high 7.18%
GPT-6 Sol max 6.25%
Claude Fable 5 high 6.22%
Claude Opus 5 max 6.18%
Tencent Hy4 5.67%
GPT-6.1 Terra medium 4.44%
Kimi K3.1 Pro max 3.95%
GPT-5.6 Sol xhigh 3.72%
Inkling 2 Small 2.51%
Grok 4.7 xhigh 1.78%
Muse Ember max 1.73%
MiMo-V2.6 Pro 1.50%
Gemini 3.8 Flash high 1.44%
Muse Spark 1.3 max 1.20%
Claude Sonnet 5 high 1.14%
DeepSeek V4.1 Flash max 1.13%
Muse Wonder max 0.95%
GPT-6 Luna max 0.79%
GLM-5.3 max 0.03%
Kimi K3 max -0.24%
Tencent Hy4 preview -0.26%
Claude Haiku 5.5 high -0.37%
GLM-5.2 max -0.49%
MiMo-V2.6 Flash -0.64%
Qwen 3.8 Max -0.87%
Grok 4.6 xhigh -0.92%
GPT-5.6 Terra xhigh -2.72%
GPT-6 Luna medium -2.89%
Gemini 3.7 Flash high -3.06%
MiniMax-M3.1-Flash -3.52%
GLM-5.3-Flash -3.70%
Step 5 Preview -4.65%
Qwen3.8-Flash-Next -5.18%
Qwen 3.8 27B -5.63%
Gemini 3.1 Pro -5.64%
Mistral Large 4 -6.81%
GPT-6 Luna low -7.18%
Beam 1.0 -7.40%
Qwen3.7-Max -8.01%
Tencent Hy3 -8.70%
Muse Spark 1.1 -9.01%
Qwen3.7-Plus -9.02%
Muse Spark 1.2 xhigh -9.08%
MiniMax-M3 -11.18%
Mistral Medium 3.5 -14.47%
Nemotron 3 Ultra -16.40%
Nemotron 3.5 Super VL nvfp4 -16.73%
Inkling -18.28%
Inkling-Small -18.70%
Solar Pro 4 -23.93%

5 models have held number one on the official leaderboard since its release in June 2026 5 models have held number one with claims included since its release in June 2026

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

Google DeepMind Anthropic
Own leaderboard 5 leaders

Source: Official Agent Arena leaderboard

Scores over time

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

Plot with other benchmarks
On the benchmark’s own leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Fable 5.2 Claude Opus 5.5 Claude Fable 5 Gemini 3.1 Pro 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Fable 5.2 Claude Opus 5.5

Sources: Official Agent Arena leaderboard and labs’ model cards

One lab put it in a launch post in the last six months

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

Qwen
DateLaunch postScoreAs the post writes it
Nov 15, 2026 Qwen4-Max 7.5 Real-world agent tasks, AgentArena (Net Improvement), cited from Arena

One system card reports it

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
Jul 27, 2026 Kimi K3 technical report 9.1 Agent Arena, rank 4 of 37 (Table 5) Other

One lab claim next to the leaderboard

Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. 1 of them match, which usually means one side copied the other. Every benchmark.

Model Lab says Board says Gap
Qwen4-Max 7.50% Real-world agent tasks, AgentArena (Net Improvement), cited from Arena 7.46% Leaderboard, xhigh Same

More in Tool use

All 265 in Tool use
  • AutomationBench In use Agent completes end-to-end business workflows inside a simulated company exposing dozens of REST APIs. Graded on resulting state. Gemini 4 Neon 56.3%
  • MCP Atlas In use Multi-step workflows over real MCP servers. Scores tool-use competency on realistic tasks. Gemini 4 Argon 90.7%
  • BFCL V4 In use Models call functions and tools, including multi-turn agentic tasks with web search and memory. Checked against expected calls and outcomes. Qwen3.7-Max 75%
  • τ³-bench Banking In use Agents act as banking customer-service reps over many turns with a simulated user, following policy. Pass^1 on the final database state. Qwen 3.8 Max 55.1%
  • Toolathlon-Verified In use 108 long-horizon tasks across 32 apps through about 600 MCP tools. Execution-based checkers inspect final environment state. GLM-5.3-Flash 78.4%
  • DeepSearchQA In use Questions that need multi-step search and synthesis. Final answers graded against references. Kimi K3 95%

Data audit

11 of 12 checks pass, 1 warning. Look first at the Confidence check. Likely. Some facts rest on secondary sources. Every benchmark

Catalog record 1 warning 1 pass
  • Pass Task access All tasks are public.
  • Warning Confidence Likely. Some facts rest on secondary sources.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (59 entries), launch posts (1 score) and system cards (1 score).
  • Pass Models scored 58 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from November 2026, for Qwen4-Max.
  • Pass Frontier models 9 of the top 10 have a score, led by Claude Fable 5.2 (#2).
  • Pass Headroom The best is 28.6%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arena.ai: 64 rows for models we track and 1 other entry at the last read.
  • Pass Last read November 18th, 2026.
  • Pass Read worked The last read worked.
Consistency 2 pass
  • Pass Lab claims The one pair agrees within 10 points.
  • Pass Known issues No published audit found problems.

How we check