Arena Coding

What does it measure?
Which AI's answers people preferred on coding questions (writing code, fixing bugs, explaining). It is not a test that checks whether the code actually runs — it records which answer the asker found more helpful.
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. A prompt is classified as coding automatically when the prompt or answer contains a code block or mentions a programming language.
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. Rankings can differ from benchmarks that execute the code (Terminal-Bench, SciCode, etc.).

Rank

Top score1568
Models tested108
Scoring basisHigher is better
Last updated2026.11.19
Model release date

Which AI do people prefer most for coding answers?

Arena Coding

  1. 🥇GoogleGemini 4 Carbon—1568—
  2. 🥈GoogleGemini 4 Neon—1559—
  3. 🥉GoogleGemini 4 Argon57.1%1552—
  4. #4AnthropicClaude Fable 542.4%155162.2
  5. #5MetaMuse Wonder52.0%1550—
  6. #6AnthropicClaude Opus 5.559.6%154865.3
  7. #7OpenAIGPT-6 Astra59.6%154857.3
  8. #8AnthropicClaude Opus 4.6—154749.0
  9. #9AnthropicClaude Opus 4.7—154650.7
  10. #10OpenAIGPT-6.1 Sol56.1%154556.8
  11. #11OpenAIGPT-6.1 Terra55.6%1541—
  12. #12MetaMuse Spark 1.27.1%154057.6
  13. #13XiaomiMiMo-V2.6-Pro34.8%1539—
  14. #14Moonshot AIKimi K312.6%153762.2
  15. #15MetaMuse Spark 1.333.3%153564.1
  16. #16MetaMuse Spark 1.16.1%153358.5
  17. #17OpenAIGPT-5.6 Sol39.9%153356.2
  18. #18AnthropicClaude Fable 5.155.1%153366.1
  19. #19Moonshot AIKimi K3.1 Pro67.2%1532—
  20. #20GoogleGemini 3.8 Flash19.7%153054.2
  21. #21MetaMuse Spark—1529—
  22. #22AnthropicClaude Opus 4.821.7%152850.5
  23. #22AnthropicClaude Sonnet 4.63.0%152842.6
  24. #24AnthropicClaude Opus 549.0%152865.2
  25. #25AnthropicClaude Fable 5.2—1527—
  26. #26Z.aiGLM 5.3 Flash32.8%152656.8
  27. #27DeepSeekDeepSeek V4.1 Flash26.8%152577.3
  28. #28Z.aiGLM 5.341.9%152460.9
  29. #29AlibabaQwen3.7 Max1.5%152443.6
  30. #30AnthropicClaude Opus 4.5—152339.7
  31. #30OpenAIGPT-6.2 Astra—1523—
  32. #32OpenAIGPT-5.4—152153.8
  33. #33GoogleGemini 3.7 Flash13.6%152158.3
  34. #33GoogleGemini 3.1 Pro4.0%152144.1
  35. #35XiaomiMiMo V2.5 Pro0.0%1520—
  36. #36AnthropicClaude Sonnet 514.1%151959.4
  37. #37OpenAIGPT-6 Sol43.9%151952.9
  38. #38GoogleGemini 3.6 Flash7.1%151843.4
  39. #39OpenAIGPT-5.6 Terra35.4%151855.0
  40. #40AnthropicClaude Sonnet 5.563.6%151739.3
  41. #41Moonshot AIKimi K2.60.5%151646.9
  42. #42Z.aiGLM 5.21.0%151651.8
  43. #43SpaceXSIGrok 4.510.6%151456.5
  44. #44ByteDanceDola Seed 2.0 Pro—1514—
  45. #45GoogleGemini 4 Hydrogen—1514—
  46. #46Z.aiGLM-5.12.0%1512—
  47. #46AnthropicClaude Sonnet 4.50.0%1512—
  48. #48AnthropicClaude Opus 4.1—1511—
  49. #49SpaceXSIGrok 4.20 (Reasoning)—1511—
  50. #50AlibabaQwen3.6 Max—1510—
  51. #51OpenAIGPT-5.514.6%151054.0
  52. #52XiaomiMiMo-V2.6-Flash22.7%1510—
  53. #53SpaceXSIGrok 4.621.2%150857.0
  54. #54SpaceXSIGrok 4.20—1508—
  55. #55GoogleGemini 3.5 Flash6.6%150649.0
  56. #56OpenAIGPT-6.2 Sol—1506—
  57. #57GoogleGemini 3 Flash—1505—
  58. #58StepfunStep-533.3%1504—
  59. #59OpenAIGPT-6 Luna12.6%150451.2
  60. #60OpenAIGPT-5.6 Luna11.6%150348.4
  61. #60XiaomiMiMo V2 Pro—1503—
  62. #62GoogleGemma 4 31B0.0%1502—
  63. #63Moonshot AIKimi K2.5—1502—
  64. #64GoogleGemini 3.5 Flash-Lite1.0%150245.3
  65. #65DeepSeekDeepSeek V4 Pro14.1%150142.6
  66. #65MeituanLongcat Flash Chat—1501—
  67. #67AlibabaQwen3.7 Plus1.0%1500—
  68. #68Z.aiGLM-5—1498—
  69. #69AnthropicClaude Opus 4—1498—
  70. #70AlibabaQwen3.6 Plus—149641.4
  71. #71OpenAIGPT-5.4 Mini2.0%149541.7
  72. #72MiniMaxMiniMax M32.0%149240.7
  73. #73XiaomiMiMo V2.50.0%1491—
  74. #74SpaceXSIGrok 4.725.8%149154.0
  75. #75AlibabaQwen3.5 397B A17B0.0%1491—
  76. #76BaiduERNIE 5.0 Thinking—1490—
  77. #77AnthropicClaude Haiku 5.532.8%148751.4
  78. #78SpaceXSIGrok 4.30.0%148718.5
  79. #79Mistral AIMistral Large 426.8%148657.2
  80. #80InklingInkling 2 Small—1485—
  81. #81DeepSeekDeepSeek V4 Flash12.1%148437.6
  82. #82AnthropicClaude Haiku 4.50.0%1482—
  83. #83MiniMaxMiniMax M2.70.0%1480—
  84. #84Mistral AIMistral Medium 3.50.0%1478—
  85. #85AnthropicClaude Sonnet 4—1475—
  86. #86NVIDIANemotron 3 Ultra0.5%147438.7
  87. #87OpenAIGPT-5—1472—
  88. #88DeepSeekDeepSeek V3.2—1470—
  89. #89GoogleGemini 2.5 Pro0.0%1464—
  90. #90SpaceXSIGrok 4.1 Fast (Reasoning)—1461—
  91. #91UpstageSolar Pro 40.0%1461—
  92. #92OpenAIGPT-5.4 Nano0.5%146046.8
  93. #93GoogleGemini 3.1 Flash Lite0.5%1458—
  94. #94OpenAIGPT-4.1—1456—
  95. #95TencentHy30.5%1452—
  96. #96MiniMaxMiniMax M2.5—1446—
  97. #97OpenAIGPT-5 Mini0.0%1432—
  98. #98GoogleGemini 2.5 Flash—1424—
  99. #99Arcee AITrinity Large Thinking0.5%1412—
  100. #100NVIDIANemotron 3 Super0.0%1407—
  101. #101GoogleGemini 2.5 Flash Lite—1398—
  102. #102AmazonNova 2 Lite—1393—
  103. #103OpenAIGPT OSS 120B0.0%1389—
  104. #104OpenAIGPT-5 Nano—1386—
  105. #105MetaLlama 4 Maverick0.0%1372—
  106. #106MetaLlama 4 Scout0.0%1362—
  107. #107NVIDIANemotron 3 Nano 30B A3B0.0%1361—
  108. #108IBMGranite 4.1 8B—1354—
Half of models ≤ 1507Measured 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
    1489

45 of the 64 AIs released in the last 60 days have no score here yet · Microsoft-Decision-1.1, d1.2, Qwen4 Max and 42 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.