Benchmark Almanac

Agent Arena

Claude Fable 5.2 30.4% leads since November 2026
Only in system cards

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.

Released
June 2026
Kind
General assistants
Level
Everyday
Scored
Ranked by votes
Answers
Agent task
Tasks
All public
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last in a system card
July 2026
Last model added
November 2026
Scores refreshed
November 12th, 2026
57
Entries on its own leaderboard, 56 models we track
1
System card reports it, from 1 lab

Claude Fable 5.2 leads with 30.4%

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

57 entries. Scroll sideways for the rest.

  1. 30.4 Claude Fable 5.2
  2. 19.5 Claude Opus 5.5
  3. 17.0 Gemini 4 Carbon
  4. 15.9 Claude Sonnet 5.5
  5. 14.8 Claude Fable 5.1
  6. 14.5 GPT-6 Astra
  7. 12.5 GPT-6.1 Sol
  8. 12.0 GPT-6.1 Terra
  9. 10.7 Gemini 4 Argon
  10. 8.96 Muse Wonder Full
  11. 7.30 Claude Fable 5
  12. 7.23 Claude Opus 5
  13. 6.91 GPT-6 Sol
  14. 5.26 Tencent Hy4
  15. 4.89 Kimi K3.1 Pro
  16. 4.63 GPT-5.6 Sol
  17. 2.56 Grok 4.7
  18. 2.51 Inkling 2 Small
  19. 2.40 MiMo-V2.6 Pro
  20. 2.30 Gemini 3.8 Flash
  21. 2.19 Muse Spark 1.3
  22. 2.00 DeepSeek V4.1 Flash
  23. 2.00 Claude Sonnet 5
  24. 1.77 Muse Wonder
  25. 1.67 Muse Ember
  26. 1.56 GPT-6 Luna
  27. 0.99 GLM-5.3
  28. 0.55 Kimi K3
  29. 0.47 Claude Haiku 5.5
  30. 0.47 Tencent Hy4 preview
  31. 0.23 GLM-5.2
  32. 0.11 Qwen 3.8 Max
  33. -0.03 MiMo-V2.6 Flash
  34. -0.19 Grok 4.6
  35. -1.85 GPT-5.6 Terra
  36. -2.20 Gemini 3.7 Flash
  37. -2.70 MiniMax-M3.1-Flash
  38. -2.95 GLM-5.3-Flash
  39. -3.74 Step 5 Preview
  40. -4.50 Gemini 3.1 Pro
  41. -4.61 Qwen3.8-Flash-Next
  42. -4.76 Qwen 3.8 27B
  43. -5.44 Mistral Large 4
  44. -5.91 Beam 1.0
  45. -7.15 Qwen3.7-Max
  46. -7.56 Tencent Hy3
  47. -7.99 Qwen3.7-Plus
  48. -8.01 Muse Spark 1.1
  49. -8.24 Muse Spark 1.2
  50. -10.46 MiniMax-M3
  51. -10.74 Spark-X2.6
  52. -13.73 Mistral Medium 3.5
  53. -15.60 Nemotron 3 Ultra
  54. -15.99 Nemotron 3.5 Super VL
  55. -17.32 Inkling
  56. -18.06 Inkling-Small
  57. -22.60 Solar Pro 4
  1. 1 Claude Fable 5.2 max 30.4%
  2. 2 Claude Opus 5.5 max 19.5%
  3. 3 Gemini 4 Carbon high 17%
  4. 4 Claude Sonnet 5.5 max 15.9%
  5. 5 Claude Fable 5.1 max 14.8%
  6. 6 GPT-6 Astra max 14.5%
  7. 7 GPT-6.1 Sol max 12.5%
  8. 8 GPT-6.1 Terra max 12%
  9. 9 Gemini 4 Argon high 10.7%
  10. 10 Muse Wonder Full max 8.96%
  11. 11 Claude Fable 5 high 7.30%
  12. 12 Claude Opus 5 max 7.23%
  13. 13 GPT-6 Sol max 6.91%
  14. 14 Tencent Hy4 5.26%
  15. 15 Kimi K3.1 Pro max 4.89%
  16. 16 GPT-5.6 Sol xhigh 4.63%
  17. 17 Grok 4.7 xhigh 2.56%
  18. 18 Inkling 2 Small 2.51%
  19. 19 MiMo-V2.6 Pro 2.40%
  20. 20 Gemini 3.8 Flash high 2.30%
  21. 21 Muse Spark 1.3 max 2.19%
  22. 22 DeepSeek V4.1 Flash max 2.00%
  23. 23 Claude Sonnet 5 high 2.00%
  24. 24 Muse Wonder max 1.77%
  25. 25 Muse Ember max 1.67%
  26. 26 GPT-6 Luna max 1.56%
  27. 27 GLM-5.3 max 0.99%
  28. 28 Kimi K3 max 0.55%
  29. 29 Claude Haiku 5.5 high 0.47%
  30. 30 Tencent Hy4 preview 0.47%
  31. 31 GLM-5.2 max 0.23%
  32. 32 Qwen 3.8 Max 0.11%
  33. 33 MiMo-V2.6 Flash -0.03%
  34. 34 Grok 4.6 xhigh -0.19%
  35. 35 GPT-5.6 Terra xhigh -1.85%
  36. 36 Gemini 3.7 Flash high -2.20%
  37. 37 MiniMax-M3.1-Flash -2.70%
  38. 38 GLM-5.3-Flash -2.95%
  39. 39 Step 5 Preview -3.74%
  40. 40 Gemini 3.1 Pro -4.50%
  41. 41 Qwen3.8-Flash-Next -4.61%
  42. 42 Qwen 3.8 27B -4.76%
  43. 43 Mistral Large 4 -5.44%
  44. 44 Beam 1.0 -5.91%
  45. 45 Qwen3.7-Max -7.15%
  46. 46 Tencent Hy3 -7.56%
  47. 47 Qwen3.7-Plus -7.99%
  48. 48 Muse Spark 1.1 -8.01%
  49. 49 Muse Spark 1.2 xhigh -8.24%
  50. 50 MiniMax-M3 -10.46%
  51. 51 Spark-X2.6 -10.74%
  52. 52 Mistral Medium 3.5 -13.73%
  53. 53 Nemotron 3 Ultra -15.60%
  54. 54 Nemotron 3.5 Super VL nvfp4 -15.99%
  55. 55 Inkling -17.32%
  56. 56 Inkling-Small -18.06%
  57. 57 Solar Pro 4 -22.60%

Source: Official Agent Arena leaderboard

#EntryScoreSettingDateSource
1 Claude Fable 5.2 30.4% max Nov 2026 Official leaderboard
2 Claude Opus 5.5 19.5% max Sep 2026 Official leaderboard
3 Gemini 4 Carbon 17% high Oct 2026 Official leaderboard
4 Claude Sonnet 5.5 15.9% max Sep 2026 Official leaderboard
5 Claude Fable 5.1 14.8% max Sep 2026 Official leaderboard
6 GPT-6 Astra 14.5% max Sep 2026 Official leaderboard
7 GPT-6.1 Sol 12.5% max Sep 2026 Official leaderboard
8 GPT-6.1 Terra 12% max Oct 2026 Official leaderboard
9 Gemini 4 Argon 10.7% high Sep 2026 Official leaderboard
10 Muse Wonder Full 8.96% max Oct 2026 Official leaderboard
11 Claude Fable 5 7.30% high Jun 2026 Official leaderboard
12 Claude Opus 5 7.23% max Jul 2026 Official leaderboard
13 GPT-6 Sol 6.91% max Sep 2026 Official leaderboard
14 Tencent Hy4 5.26% Nov 2026 Official leaderboard
15 Kimi K3.1 Pro 4.89% max Oct 2026 Official leaderboard
16 GPT-5.6 Sol 4.63% xhigh Jul 2026 Official leaderboard
17 Grok 4.7 2.56% xhigh Sep 2026 Official leaderboard
18 Inkling 2 Small 2.51% Nov 2026 Official leaderboard
19 MiMo-V2.6 Pro 2.40% Sep 2026 Official leaderboard
20 Gemini 3.8 Flash 2.30% high Sep 2026 Official leaderboard
21 Muse Spark 1.3 2.19% max Sep 2026 Official leaderboard
22 DeepSeek V4.1 Flash 2.00% max Sep 2026 Official leaderboard
23 Claude Sonnet 5 2.00% high Jun 2026 Official leaderboard
24 Muse Wonder 1.77% max Oct 2026 Official leaderboard
25 Muse Ember 1.67% max Nov 2026 Official leaderboard
26 GPT-6 Luna 1.56% max Sep 2026 Official leaderboard
27 GLM-5.3 0.99% max Aug 2026 Official leaderboard
28 Kimi K3 0.55% max Jul 2026 Official leaderboard
29 Claude Haiku 5.5 0.47% high Oct 2026 Official leaderboard
30 Tencent Hy4 preview 0.47% Aug 2026 Official leaderboard
31 GLM-5.2 0.23% max Jun 2026 Official leaderboard
32 Qwen 3.8 Max 0.11% Aug 2026 Official leaderboard
33 MiMo-V2.6 Flash -0.03% Sep 2026 Official leaderboard
34 Grok 4.6 -0.19% xhigh Aug 2026 Official leaderboard
35 GPT-5.6 Terra -1.85% xhigh Jul 2026 Official leaderboard
36 Gemini 3.7 Flash -2.20% high Aug 2026 Official leaderboard
37 MiniMax-M3.1-Flash -2.70% Oct 2026 Official leaderboard
38 GLM-5.3-Flash -2.95% Aug 2026 Official leaderboard
39 Step 5 Preview -3.74% Sep 2026 Official leaderboard
40 Gemini 3.1 Pro -4.50% Feb 2026 Official leaderboard
41 Qwen3.8-Flash-Next -4.61% Aug 2026 Official leaderboard
42 Qwen 3.8 27B -4.76% Aug 2026 Official leaderboard
43 Mistral Large 4 -5.44% Oct 2026 Official leaderboard
44 Beam 1.0 -5.91% Oct 2026 Official leaderboard
45 Qwen3.7-Max -7.15% May 2026 Official leaderboard
46 Tencent Hy3 -7.56% Jul 2026 Official leaderboard
47 Qwen3.7-Plus -7.99% Jun 2026 Official leaderboard
48 Muse Spark 1.1 -8.01% Jul 2026 Official leaderboard
49 Muse Spark 1.2 -8.24% xhigh Aug 2026 Official leaderboard
50 MiniMax-M3 -10.46% Jun 2026 Official leaderboard
51 Spark-X2.6 -10.74% Nov 2026 Official leaderboard
52 Mistral Medium 3.5 -13.73% Apr 2026 Official leaderboard
53 Nemotron 3 Ultra -15.60% Jun 2026 Official leaderboard
54 Nemotron 3.5 Super VL -15.99% nvfp4 Oct 2026 Official leaderboard
55 Inkling -17.32% Jul 2026 Official leaderboard
56 Inkling-Small -18.06% Jul 2026 Official leaderboard
57 Solar Pro 4 -22.60% Aug 2026 Official leaderboard

5 models have held number one on the official leaderboard 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
Claude Fable 5.2 30.4%

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.

On the benchmark’s own leaderboard 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 5.1 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

Source: Official Agent Arena leaderboard

Labs report it only in system cards

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

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

More in Tool use

All 144 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 Carbon 53.9%
  • 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%