Arena Math

What does it measure?
Preference on maths questions (calculation, worked solutions, proofs). Because people choose rather than checking against an answer key, solutions that are easy to read have an advantage.
Who evaluates it, and how?
On Arena Intelligence, a user enters a question, two anonymous AIs answer, and the user votes for the better one. Over 82 million votes are aggregated with a statistical model (Bradley-Terry) into an Elo-style score, and this table takes the 'style-controlled' version from the official dataset, which statistically removes the bias toward long, nicely formatted answers. An AI classifier selects prompts that actively apply mathematical concepts — calculations and derivations.
Reading the score
This is a head-to-head record, not an exam score (accuracy %). It is computed from the probability of being preferred over other AIs, so there is no maximum, and when a new AI enters, existing scores shift slightly. Typical values are 1,000–1,500; a 100-point gap means the leader is preferred about 64% of the time, a 200-point gap about 76%. Correctness is not counted — a wrong but friendly answer can win — so if accuracy matters, also look at accuracy tests (GPQA, AIME, etc.). Gaps of a few dozen points are effectively the same tier even if the rank differs. It measures something different from accuracy tests (AIME, MATH-500): not whether the answer was right, but whether people liked it.

Rank

Top score1535
Models tested104
Scoring basisHigher is better
Last updated2026.11.19
Model release date

Which AI gives the math answers people prefer most?

Arena Math

  1. 🥇GoogleGemini 4 Argon57.1%27.1%1535
  2. 🥈GoogleGemini 4 Neon——1535
  3. 🥉AnthropicClaude Fable 5.159.1%31.1%1530
  4. #4AnthropicClaude Fable 555.5%28.6%1521
  5. #5GoogleGemini 3.8 Flash47.8%18.3%1519
  6. #6AnthropicClaude Opus 5.561.4%31.7%1517
  7. #7AnthropicClaude Opus 554.9%29.1%1516
  8. #8Z.aiGLM 5.3 Flash39.9%15.4%1509
  9. #9AnthropicClaude Fable 5.2—35.7%1507
  10. #10MetaMuse Wonder50.1%—1504
  11. #11GoogleGemini 4 Hydrogen——1504
  12. #12Z.aiGLM 5.342.3%19.1%1504
  13. #13AnthropicClaude Opus 4.639.9%—1504
  14. #14Moonshot AIKimi K346.9%23.4%1502
  15. #15MetaMuse Spark 1.348.7%26.0%1502
  16. #16GoogleGemini 3.7 Flash47.9%14.3%1501
  17. #16GoogleGemini 3.6 Flash40.8%10.6%1501
  18. #18OpenAIGPT-5.545.8%27.1%1500
  19. #19OpenAIGPT-6.1 Sol52.9%31.7%1499
  20. #20AnthropicClaude Sonnet 5.555.0%31.4%1497
  21. #21DeepSeekDeepSeek V4.1 Flash39.2%14.3%1496
  22. #22OpenAIGPT-5.6 Sol49.5%32.3%1494
  23. #23MetaMuse Spark 1.146.2%15.1%1494
  24. #24OpenAIGPT-5.443.7%23.4%1491
  25. #25AnthropicClaude Opus 4.742.3%12.0%1490
  26. #26GoogleGemini 3.1 Pro47.0%17.7%1488
  27. #27Z.aiGLM 5.241.1%20.9%1488
  28. #28MetaMuse Spark 1.245.5%17.7%1483
  29. #29GoogleGemini 3.5 Flash42.7%13.1%1482
  30. #30XiaomiMiMo-V2.6-Pro49.4%26.6%1482
  31. #31AlibabaQwen3.7 Max40.5%13.4%1481
  32. #32OpenAIGPT-6.1 Terra46.2%28.0%1479
  33. #33Moonshot AIKimi K2.637.5%8.0%1479
  34. #34XiaomiMiMo V2.5 Pro35.7%4.0%1479
  35. #35OpenAIGPT-5.6 Terra42.9%30.0%1478
  36. #36OpenAIGPT-6 Astra54.7%31.7%1477
  37. #37Moonshot AIKimi K3.1 Pro52.5%24.6%1477
  38. #38OpenAIGPT-5.6 Luna39.5%20.6%1476
  39. #39AnthropicClaude Opus 4.848.7%20.9%1475
  40. #40GoogleGemini 3 Flash36.6%—1475
  41. #41Z.aiGLM-5.130.1%4.6%1474
  42. #42StepfunStep-546.5%20.9%1474
  43. #43AlibabaQwen3.6 Max30.8%—1472
  44. #44AnthropicClaude Sonnet 541.3%16.9%1472
  45. #45SpaceXSIGrok 4.542.7%15.4%1471
  46. #46Moonshot AIKimi K2.530.7%3.1%1470
  47. #47MetaMuse Spark40.7%11.3%1468
  48. #48GoogleGemma 4 31B23.6%1.4%1466
  49. #49AlibabaQwen3.7 Plus35.6%9.1%1465
  50. #50SpaceXSIGrok 4.20 (Reasoning)34.5%—1464
  51. #51XiaomiMiMo-V2.6-Flash35.1%12.0%1463
  52. #52AnthropicClaude Opus 4.530.1%—1463
  53. #53AnthropicClaude Sonnet 4.633.6%3.1%1461
  54. #54AlibabaQwen3.6 Plus27.8%2.9%1451
  55. #55AlibabaQwen3.5 397B A17B29.0%0.9%1451
  56. #56XiaomiMiMo V2 Pro30.4%—1450
  57. #57ByteDanceDola Seed 2.0 Pro——1449
  58. #58SpaceXSIGrok 4.2027.9%—1449
  59. #59OpenAIGPT-6 Luna38.5%19.4%1449
  60. #60Mistral AIMistral Large 435.0%11.1%1448
  61. #61AnthropicClaude Haiku 5.544.4%22.6%1447
  62. #62SpaceXSIGrok 4.644.1%19.7%1445
  63. #63OpenAIGPT-6 Sol47.9%30.9%1444
  64. #64AnthropicClaude Opus 4.112.5%—1444
  65. #65SpaceXSIGrok 4.743.1%18.0%1443
  66. #66DeepSeekDeepSeek V4 Pro41.0%18.6%1443
  67. #67Z.aiGLM-529.3%—1443
  68. #68NVIDIANemotron 3 Ultra28.4%3.1%1442
  69. #69GoogleGemini 3.5 Flash-Lite18.8%0.0%1442
  70. #70GoogleGemini 2.5 Pro22.5%2.0%1439
  71. #71OpenAIGPT-5.4 Mini28.1%10.0%1439
  72. #72XiaomiMiMo V2.527.2%3.7%1437
  73. #73GoogleGemini 3.1 Flash Lite17.2%1.1%1436
  74. #74OpenAIGPT-528.5%12.6%1435
  75. #75MeituanLongcat Flash Chat5.8%—1434
  76. #76BaiduERNIE 5.0 Thinking13.3%—1434
  77. #77TencentHy333.5%—1429
  78. #78Mistral AIMistral Medium 3.513.8%0.0%1427
  79. #79AnthropicClaude Sonnet 4.517.8%1.1%1426
  80. #80DeepSeekDeepSeek V3.224.6%—1426
  81. #81DeepSeekDeepSeek V4 Flash38.6%16.6%1424
  82. #82OpenAIGPT-5.4 Nano28.3%9.3%1424
  83. #83MiniMaxMiniMax M339.0%3.7%1423
  84. #84MiniMaxMiniMax M2.729.6%0.6%1423
  85. #85AnthropicClaude Opus 412.3%0.3%1421
  86. #86SpaceXSIGrok 4.1 Fast (Reasoning)19.3%—1420
  87. #87SpaceXSIGrok 4.337.2%8.0%1419
  88. #88UpstageSolar Pro 429.8%—1416
  89. #89OpenAIGPT-5 Mini21.5%0.0%1407
  90. #90GoogleGemini 2.5 Flash12.1%1.1%1405
  91. #91AnthropicClaude Sonnet 410.7%0.3%1404
  92. #92AnthropicClaude Haiku 4.510.4%0.0%1402
  93. #93MiniMaxMiniMax M2.520.5%—1392
  94. #94Arcee AITrinity Large Thinking15.8%0.9%1383
  95. #95OpenAIGPT OSS 120B19.6%1.1%1381
  96. #96OpenAIGPT-4.14.2%—1375
  97. #97NVIDIANemotron 3 Super20.8%3.1%1370
  98. #98GoogleGemini 2.5 Flash Lite7.0%—1361
  99. #99OpenAIGPT-5 Nano9.5%—1348
  100. #100NVIDIANemotron 3 Nano 30B A3B11.4%—1347
  101. #101AmazonNova 2 Lite11.6%—1329
  102. #102IBMGranite 4.1 8B3.8%—1318
  103. #103MetaLlama 4 Maverick4.9%0.0%1317
  104. #104MetaLlama 4 Scout3.8%0.0%1310
Half of models ≤ 1462Measured by: User preference votes

Scores from an earlier edition

These scores are not in the latest leaderboard (2026.11.19) — the model was dropped by the source or has not been refreshed yet. The last score we received is shown for reference, unranked.

  1. Z.ai2026.04.01 · as of 2026.11.09
    1437

49 of the 64 AIs released in the last 60 days have no score here yet · GPT-6.2 Astra, Microsoft-Decision-1.1, GPT-6.2 Sol and 46 more

It’s a preference vote, not a correctness check. A wrong but friendly answer can win.

Which questions got asked depends on who voted — they may differ from your use.

Scores labelled 'Self-reported' were published by the AI maker itself, not measured by a third party.