Skip to content
Model Index
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
LabsMethodology
  1. Benchmarks
  2. Overall

Arena · Text

Human preferenceArena · updated Oct 2, 2026

What it measures

Which answer people prefer when two anonymous models respond to the same prompt, aggregated over millions of blind votes into an Elo-style rating (style control on).

How to read it

Measures how much people like the answers, not whether they are correct. A 20-point lead is only about a 53% win rate, and new models with few votes carry wide error bars.

Open at Arena
Best
1525
Gemini 4 Argon
Median
1464
DeepSeek V4 Pro

Results

33 of 41 tracked models have a result. This scale has no meaningful zero, so bars run from the lowest to the highest tracked score.

Filter in explorer
  1. 1
    Gemini 4 ArgonLimited

    Google DeepMind · High effort · 4,932 votes

    1525 ±9
  2. 2
    Claude Opus 5.5

    Anthropic · High effort · 4,552 votes

    1504 ±9
  3. 3
    Claude Fable 5.1

    Anthropic · Max effort · 11,800 votes

    1501 ±6
  4. 4
    Gemini 3.8 Flash

    Google DeepMind · High effort · 26,298 votes

    1495 ±5
  5. 5
    Muse Spark 1.3

    Meta · Max effort · 12,343 votes

    1494 ±6
  6. 6
    Kimi K3Open

    Moonshot AI · Max effort · 28,280 votes

    1488 ±5
  7. 7
    Gemini 3.1 Pro

    Google DeepMind · 121,806 votes

    1487 ±3
  8. 8
    GPT-6.1 Sol

    OpenAI · Max effort · 3,071 votes

    1483 ±11
  9. 9
    Qwen3.8-Max

    Alibaba · Max effort · 23,353 votes

    1482 ±5
  10. 10
    MiMo-V2.6-ProOpen

    Xiaomi · 4,056 votes

    1480 ±9
  11. 11
    GLM-5.3Open

    Z.ai · Max effort · 17,857 votes

    1478 ±6
  12. 12
    GPT-6 Astra

    OpenAI · Max effort · 9,156 votes

    1477 ±7
  13. 13
    DeepSeek V4.1 FlashOpen

    DeepSeek · Max effort · 8,728 votes

    1474 ±7
  14. 14
    GLM-5.3-FlashOpen

    Z.ai · 22,971 votes

    1473 ±5
  15. 15
    Claude Sonnet 5.5

    Anthropic · xHigh effort · 3,145 votes

    1471 ±10
  16. 16
    GPT-5.6 Terra

    OpenAI · xHigh effort · 37,663 votes

    1465 ±4
  17. 17
    DeepSeek V4 ProOpen

    DeepSeek · 9,780 votes

    1464 ±7
  18. 18
    Hy3Open

    Tencent · 10,476 votes

    1456 ±7
  19. 19
    Gemini 3.5 Flash-Lite

    Google DeepMind · 35,763 votes

    1455 ±4
  20. 20
    Step 5 Preview

    StepFun · 2,701 votes

    1454 ±11
  21. 21
    Gemma 4 31BOpen

    Google DeepMind · 6,133 votes

    1453 ±7
  22. 22
    MiMo-V2.6-FlashOpen

    Xiaomi · 6,053 votes

    1452 ±8
  23. 23
    GPT-6 Luna

    OpenAI · Max effort · 10,014 votes

    1443 ±7
  24. 24
    Grok 4.7

    SpaceXAI · xHigh effort · 6,405 votes

    1442 ±8
  25. 25
    TM
    InklingOpen

    Thinking Machines Lab · 35,204 votes

    1441 ±5
  26. 26
    MiniMax-M3Open

    MiniMax · 58,757 votes

    1440 ±4
  27. 27
    Qwen3.8-27BOpen

    Alibaba · 18,462 votes

    1438 ±5
  28. 28
    Mistral Medium 3.5Open

    Mistral AI · 11,757 votes

    1427 ±6
  29. 29
    Nemotron 3 UltraOpen

    NVIDIA · 11,515 votes

    1426 ±7
  30. 30
    Muse Glimmer 30BOpen

    Meta · 3,893 votes

    1425 ±10
  31. 31
    TM
    Inkling-SmallOpen

    Thinking Machines Lab · 27,070 votes

    1405 ±5
  32. 32
    gpt-oss-120bOpen

    OpenAI · 30,018 votes

    1352 ±4
  33. 33
    Nova 2 Lite

    Amazon · 12,392 votes

    1335 ±6
8 tracked models without a result

Claude Haiku 5.5, Command A+, Hy4 preview, Kimi K2.7 Code, MAI-Thinking-1, Mistral Large 4, Mistral Small 4, Qwen3.8-Flash.

Missing means the evaluator has not published a run, not that the model would score poorly.

Model Index

An independent snapshot of frontier SI models and how they perform on third-party benchmarks.

Data as of Oct 8, 2026

Explore

  • All models
  • Compare
  • Benchmarks
  • Model finder
  • Labs
  • Methodology & sources

Sources

Benchmark results come from Artificial Analysis, Arena, Vals AI and ARC Prize. Specifications and prices come from each lab's own documentation.

Not affiliated with any lab or evaluator. Names and logos are trademarks of their respective owners and are used only to identify them. Scores change often — check the linked source before making a decision that depends on one.

  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder