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

Arena · WebDev

Human preferenceArena · updated Oct 7, 2026

What it measures

Blind head-to-head votes on working web apps that two models build from the same prompt.

How to read it

Captures front-end taste and whether the result runs. It says little about back-end, large-repo or long-running work.

Open at Arena
Best
1814
Claude Opus 5.5
Median
1592
Qwen3.8-27B

Results

34 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
    Claude Opus 5.5

    Anthropic · Max effort · 2,229 votes

    1814 ±16
  2. 2
    GPT-6 Astra

    OpenAI · Max effort · 5,057 votes

    1788 ±11
  3. 3
    Claude Sonnet 5.5

    Anthropic · xHigh effort · 2,150 votes

    1773 ±15
  4. 4
    GPT-6.1 Sol

    OpenAI · Max effort · 1,965 votes

    1757 ±15
  5. 5
    Claude Fable 5.1

    Anthropic · Max effort · 6,521 votes

    1745 ±9
  6. 6
    Gemini 4 ArgonLimited

    Google DeepMind · High effort · 1,772 votes

    1678 ±15
  7. 7
    Qwen3.8-Max

    Alibaba · Max effort · 3,450 votes

    1672 ±12
  8. 8
    Muse Spark 1.3

    Meta · Max effort · 7,678 votes

    1656 ±8
  9. 9
    Kimi K3Open

    Moonshot AI · Max effort · 17,061 votes

    1655 ±6
  10. 10
    Grok 4.7

    SpaceXAI · xHigh effort · 3,734 votes

    1638 ±11
  11. 11
    MiMo-V2.6-FlashOpen

    Xiaomi · 3,851 votes

    1636 ±11
  12. 12
    Hy4 previewOpen

    Tencent · 4,798 votes

    1632 ±10
  13. 12
    Qwen3.8-FlashOpen

    Alibaba · 6,711 votes

    1632 ±9
  14. 14
    MiMo-V2.6-ProOpen

    Xiaomi · 3,415 votes

    1630 ±11
  15. 15
    GLM-5.3Open

    Z.ai · Max effort · 8,195 votes

    1622 ±8
  16. 16
    DeepSeek V4.1 FlashOpen

    DeepSeek · Max effort · 3,547 votes

    1619 ±11
  17. 17
    GLM-5.3-FlashOpen

    Z.ai · 12,868 votes

    1610 ±7
  18. 18
    Qwen3.8-27BOpen

    Alibaba · 13,361 votes

    1592 ±7
  19. 19
    Gemini 3.8 Flash

    Google DeepMind · High effort · 8,981 votes

    1583 ±8
  20. 20
    DeepSeek V4 ProOpen

    DeepSeek · 4,744 votes

    1582 ±9
  21. 20
    GPT-6 Luna

    OpenAI · Max effort · 10,729 votes

    1582 ±7
  22. 22
    Step 5 Preview

    StepFun · High effort · 3,138 votes

    1564 ±12
  23. 23
    Mistral Large 4

    Mistral AI · 872 votes

    1534 ±21
  24. 24
    GPT-5.6 Terra

    OpenAI · 12,128 votes

    1522 ±7
  25. 25
    Hy3Open

    Tencent · 5,543 votes

    1508 ±9
  26. 26
    MiniMax-M3Open

    MiniMax · 18,086 votes

    1482 ±6
  27. 27
    Kimi K2.7 CodeOpen

    Moonshot AI · 4,986 votes

    1473 ±9
  28. 28
    Gemini 3.1 Pro

    Google DeepMind · 24,296 votes

    1447 ±5
  29. 29
    Gemini 3.5 Flash-Lite

    Google DeepMind · 242 votes

    1440 ±41
  30. 30
    TM
    InklingOpen

    Thinking Machines Lab · 14,166 votes

    1412 ±6
  31. 31
    TM
    Inkling-SmallOpen

    Thinking Machines Lab · 5,372 votes

    1409 ±9
  32. 32
    Gemma 4 31BOpen

    Google DeepMind · 13,592 votes

    1365 ±6
  33. 33
    Muse Glimmer 30BOpen

    Meta · 1,659 votes

    1355 ±16
  34. 34
    Mistral Medium 3.5Open

    Mistral AI · 2,325 votes

    1263 ±15
7 tracked models without a result

Claude Haiku 5.5, Command A+, gpt-oss-120b, MAI-Thinking-1, Mistral Small 4, Nemotron 3 Ultra, Nova 2 Lite.

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