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

Terminal-Bench 4.0

Independent testArtificial Analysis · updated Oct 8, 2026

What it measures

End-to-end software and systems tasks solved by an agent inside a real terminal, graded by a test suite (pass@1 over three repeats).

How to read it

The closest public proxy for autonomous coding-agent work. Absolute numbers depend heavily on the agent harness, so compare models only within one evaluator — these are Artificial Analysis runs.

Open at Artificial Analysis
Best
63.6%
Claude Sonnet 5.5
Median
19.7%
Gemini 3.8 Flash

Results

38 of 41 tracked models have a result.

Filter in explorer
  1. 1
    Claude Sonnet 5.5

    Anthropic · Max effort

    63.6%
  2. 2
    Claude Opus 5.5

    Anthropic · Max effort

    59.6%
  3. 3
    GPT-6 Astra

    OpenAI · Max effort

    59.1%
  4. 4
    Gemini 4 ArgonLimited

    Google DeepMind · High effort

    57.1%
  5. 5
    GPT-6.1 Sol

    OpenAI · Max effort

    56.1%
  6. 6
    Claude Fable 5.1

    Anthropic · Max effort

    52.0%
  7. 7
    GLM-5.3Open

    Z.ai · Max effort

    41.9%
  8. 8
    Qwen3.8-Max

    Alibaba

    38.9%
  9. 9
    GPT-5.6 Terra

    OpenAI · Max effort

    35.4%
  10. 10
    MiMo-V2.6-ProOpen

    Xiaomi

    34.8%
  11. 11
    Muse Spark 1.3

    Meta · Max effort

    33.3%
  12. 11
    Step 5 Preview

    StepFun

    33.3%
  13. 13
    Claude Haiku 5.5

    Anthropic · Max effort

    32.8%
  14. 13
    GLM-5.3-FlashOpen

    Z.ai

    32.8%
  15. 15
    DeepSeek V4.1 FlashOpen

    DeepSeek · Max effort

    26.8%
  16. 15
    Mistral Large 4

    Mistral AI

    26.8%
  17. 17
    Grok 4.7

    SpaceXAI · xHigh effort

    25.8%
  18. 18
    Qwen3.8-FlashOpen

    Alibaba

    25.3%
  19. 19
    MiMo-V2.6-FlashOpen

    Xiaomi

    22.7%
  20. 20
    Gemini 3.8 Flash

    Google DeepMind · High effort

    19.7%
  21. 21
    DeepSeek V4 ProOpen

    DeepSeek · Max effort

    14.1%
  22. 22
    GPT-6 Luna

    OpenAI · Max effort

    12.6%
  23. 22
    Kimi K3Open

    Moonshot AI · Max effort

    12.6%
  24. 24
    Qwen3.8-27BOpen

    Alibaba · xHigh effort

    5.6%
  25. 25
    Gemini 3.1 Pro

    Google DeepMind

    4.0%
  26. 26
    MiniMax-M3Open

    MiniMax

    2.0%
  27. 27
    Gemini 3.5 Flash-Lite

    Google DeepMind

    1.0%
  28. 27
    TM
    InklingOpen

    Thinking Machines Lab · xHigh effort

    1.0%
  29. 27
    TM
    Inkling-SmallOpen

    Thinking Machines Lab

    1.0%
  30. 27
    Kimi K2.7 CodeOpen

    Moonshot AI

    1.0%
  31. 31
    Command A+Open

    Cohere

    0.5%
  32. 31
    Hy3Open

    Tencent

    0.5%
  33. 31
    Muse Glimmer 30BOpen

    Meta · High effort

    0.5%
  34. 31
    Nemotron 3 UltraOpen

    NVIDIA · Reasoning effort

    0.5%
  35. 35
    Gemma 4 31BOpen

    Google DeepMind · Reasoning effort

    0.0%
  36. 35
    gpt-oss-120bOpen

    OpenAI · High effort

    0.0%
  37. 35
    Mistral Medium 3.5Open

    Mistral AI

    0.0%
  38. 35
    Mistral Small 4Open

    Mistral AI · Reasoning effort

    0.0%
3 tracked models without a result

Hy4 preview, MAI-Thinking-1, 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