Acropolis

Terminal-Bench 4

Terminal-Bench 4, a component of the Vals Index.

Upstream source
CodingAgentic
Independent. Run by an operator independent of the model vendors.Vals AI benchmarks by Vals AIVersion 4.0Data as of 17 Nov 2026

No filters. Showing 38 of 38 rows.

  1. 1
    AnthropicDossier

    high

    Accuracy

    Fable 5.2, Accuracy: 65.2% (scale 0.0 to 100.0), 95% confidence interval 60.6 to 69.765.2%(60.6–69.7)
    Duration per test
    3,307.2 s
    Cost per test
    $14.56
  2. 1
    AnthropicDossier

    max

    Accuracy

    Opus 5.5, Accuracy: 65.2% (scale 0.0 to 100.0), 95% confidence interval 65.2 to 65.265.2%(65.2–65.2)
    Duration per test
    3,810.9 s
    Cost per test
    $12.04
  3. 3
    AnthropicDossier

    max

    Tied #1

    Accuracy

    Sonnet 5.5, Accuracy: 64.1% (scale 0.0 to 100.0), 95% confidence interval 62.2 to 66.164.1%(62.2–66.1)
    Duration per test
    4,928.9 s
    Cost per test
    $10.75
  4. 4
    Moonshot AIDossier

    max

    Tied #1

    Accuracy

    Kimi K3.1 Pro, Accuracy: 63.6% (scale 0.0 to 100.0), 95% confidence interval 60.2 to 67.163.6%(60.2–67.1)
    Duration per test
    2,960.7 s
    Cost per test
    $4.57
  5. 5
    GoogleDossier

    high

    Tied #1

    Accuracy

    Gemini 4 Carbon, Accuracy: 63.1% (scale 0.0 to 100.0), 95% confidence interval 58.8 to 67.463.1%(58.8–67.4)
    Duration per test
    2,706.8 s
    Cost per test
    $7.45
  6. 6
    OpenAIDossier

    max

    Tied #1

    Accuracy

    GPT-6 Astra, Accuracy: 59.6% (scale 0.0 to 100.0), 95% confidence interval 51.0 to 68.259.6%(51.0–68.2)
    Duration per test
    2,142.1 s
    Cost per test
    $9.58
  7. 7

    max

    Tied #1

    Accuracy

    Muse Wonder Full 1.0, Accuracy: 58.1% (scale 0.0 to 100.0), 95% confidence interval 51.6 to 64.658.1%(51.6–64.6)
    Duration per test
    4,330.7 s
    Cost per test
    $10.53
  8. 8
    GoogleDossier

    high

    Accuracy

    Gemini 4 Argon, Accuracy: 57.6% (scale 0.0 to 100.0), 95% confidence interval 53.0 to 62.157.6%(53.0–62.1)
    Duration per test
    3,214.8 s
    Cost per test
    $17.64
  9. 9
    OpenAIDossier

    max

    Accuracy

    GPT-6.1 Sol, Accuracy: 55.1% (scale 0.0 to 100.0), 95% confidence interval 51.5 to 58.655.1%(51.5–58.6)
    Duration per test
    2,711.7 s
    Cost per test
    $1.72
  10. 10
    GoogleDossier

    xhigh

    Accuracy

    Gemini 4 Hydrogen, Accuracy: 54% (scale 0.0 to 100.0), 95% confidence interval 47.5 to 60.554%(47.5–60.5)
    Duration per test
    1,839.9 s
    Cost per test
    $7.56
  11. 11
    OpenAIDossier

    max

    Accuracy

    GPT-6.1 Terra, Accuracy: 49.5% (scale 0.0 to 100.0), 95% confidence interval 45.9 to 53.149.5%(45.9–53.1)
    Duration per test
    2,032.8 s
    Cost per test
    $2.16
  12. 11

    max

    Accuracy

    Muse Wonder 1.0, Accuracy: 49.5% (scale 0.0 to 100.0), 95% confidence interval 45.2 to 53.849.5%(45.2–53.8)
    Duration per test
    3,280.7 s
    Cost per test
    $14.66
  13. 13
    AlibabaDossier

    xhigh

    Accuracy

    Qwen4 Max, Accuracy: 48.5% (scale 0.0 to 100.0), 95% confidence interval 46.8 to 50.248.5%(46.8–50.2)
    Duration per test
    4,489.8 s
    Cost per test
    $3.90
  14. 14
    TencentDossier

    Accuracy

    Hy4, Accuracy: 39.4% (scale 0.0 to 100.0), 95% confidence interval 37.7 to 41.139.4%(37.7–41.1)
    Duration per test
    2,600.4 s
    Cost per test
    $1.28
  15. 14

    max

    Accuracy

    Muse Ember, Accuracy: 39.4% (scale 0.0 to 100.0), 95% confidence interval 34.9 to 43.939.4%(34.9–43.9)
    Duration per test
    1,581.4 s
    Cost per test
    $0.22
  16. 16

    max

    Accuracy

    GLM-5.3, Accuracy: 38.9% (scale 0.0 to 100.0), 95% confidence interval 35.3 to 42.538.9%(35.3–42.5)
    Duration per test
    4,952.9 s
    Cost per test
    $9.37
  17. 16
    AlibabaDossier

    Accuracy

    Qwen3.8 Max Internal, Accuracy: 38.9% (scale 0.0 to 100.0), 95% confidence interval 36.3 to 41.538.9%(36.3–41.5)
    Duration per test
    8,591.4 s
    Cost per test
    $12.15
  18. 18

    max

    Accuracy

    GLM-5.3 Flash, Accuracy: 37.4% (scale 0.0 to 100.0), 95% confidence interval 32.1 to 42.637.4%(32.1–42.6)
    Duration per test
    3,681.1 s
    Cost per test
    $1.22
  19. 19
    AnthropicDossier

    max

    Accuracy

    Haiku 5.5, Accuracy: 36.4% (scale 0.0 to 100.0), 95% confidence interval 31.8 to 40.936.4%(31.8–40.9)
    Duration per test
    3,116.4 s
    Cost per test
    $3.12
  20. 20
    DeepSeekDossier

    max

    Accuracy

    DeepSeek V4.1 Flash 1017, Accuracy: 31.8% (scale 0.0 to 100.0), 95% confidence interval 26.7 to 37.031.8%(26.7–37.0)
    Duration per test
    2,470.4 s
    Cost per test
    $0.54
  21. 20
    StepFunDossier

    Accuracy

    Step 5 Preview, Accuracy: 31.8% (scale 0.0 to 100.0), 95% confidence interval 26.7 to 37.031.8%(26.7–37.0)
    Duration per test
    7,067 s
    Cost per test
    $3.11
  22. 22
    XiaomiDossier

    Accuracy

    MiMo-V2.6-Pro, Accuracy: 31.3% (scale 0.0 to 100.0), 95% confidence interval 25.3 to 37.331.3%(25.3–37.3)
    Duration per test
    9,720 s
    Cost per test
    $0.46
  23. 23

    xhigh

    Accuracy

    Grok 4.7, Accuracy: 28.8% (scale 0.0 to 100.0), 95% confidence interval 24.3 to 33.328.8%(24.3–33.3)
    Duration per test
    3,007.5 s
    Cost per test
    $18.01
  24. 24
    MistralDossier

    high

    Accuracy

    Mistral Large 4 OSS, Accuracy: 26.3% (scale 0.0 to 100.0), 95% confidence interval 16.2 to 36.326.3%(16.2–36.3)
    Duration per test
    3,872.7 s
    Cost per test
    $2.73
  25. 25
    MiniMaxDossier

    max

    Accuracy

    MiniMax M3.1 Flash, Accuracy: 25.3% (scale 0.0 to 100.0), 95% confidence interval 23.3 to 27.225.3%(23.3–27.2)
    Duration per test
    3,030.6 s
    Cost per test
    $0.73
  26. 26
    XiaomiDossier

    Accuracy

    MiMo-V2.6-Flash, Accuracy: 24.2% (scale 0.0 to 100.0), 95% confidence interval 21.3 to 27.224.2%(21.3–27.2)
    Duration per test
    8,712.2 s
    Cost per test
    $0.21
  27. 27
    GoogleDossier

    high

    Accuracy

    Gemini 3.8 Flash, Accuracy: 19.7% (scale 0.0 to 100.0), 95% confidence interval 15.2 to 24.219.7%(15.2–24.2)
    Duration per test
    6,427.6 s
    Cost per test
    $8.65
  28. 28
    Moonshot AIDossier

    max

    Accuracy

    Kimi K3, Accuracy: 17.2% (scale 0.0 to 100.0), 95% confidence interval 16.2 to 18.217.2%(16.2–18.2)
    Duration per test
    11,353.5 s
    Cost per test
    $12.48
  29. 29
    DeepSeekDossier

    max

    Accuracy

    DeepSeek V4 Pro 0813, Accuracy: 14.1% (scale 0.0 to 100.0), 95% confidence interval 10.2 to 18.114.1%(10.2–18.1)
    Duration per test
    5,493.6 s
    Cost per test
    $3.31
  30. 30
    OpenAIDossier

    max

    Accuracy

    GPT-6 Luna, Accuracy: 13.6% (scale 0.0 to 100.0), 95% confidence interval 10.7 to 16.613.6%(10.7–16.6)
    Duration per test
    1,965.6 s
    Cost per test
    $0.35
  31. 31
    Thinking MachinesDossier

    Accuracy

    Inkling 2 Small, Accuracy: 12.6% (scale 0.0 to 100.0), 95% confidence interval 6.6 to 18.612.6%(6.6–18.6)
    Duration per test
    6,062.7 s
    Cost per test
    $8.22
  32. 32
    TencentDossier

    Accuracy

    Hy4 Preview, Accuracy: 8.1% (scale 0.0 to 100.0), 95% confidence interval 5.5 to 10.78.1%(5.5–10.7)
    Duration per test
    7,798.7 s
    Cost per test
    $3.07
  33. 33

    xhigh

    Accuracy

    Muse Spark Open, Accuracy: 6.6% (scale 0.0 to 100.0), 95% confidence interval 5.6 to 7.66.6%(5.6–7.6)
    Duration per test
    7,466.4 s
    Cost per test
    $7.26
  34. 34
    ReflectionDossier

    Accuracy

    Beam 1.0, Accuracy: 4% (scale 0.0 to 100.0), 95% confidence interval 1.4 to 6.74%(1.4–6.7)
    Duration per test
    1,154.6 s
    Cost per test
    Not tested
  35. 35
    InceptionDossier

    high

    Accuracy

    Mercury 3, Accuracy: 3% (scale 0.0 to 100.0), 95% confidence interval 3.0 to 3.03%(3.0–3.0)
    Duration per test
    1,987.2 s
    Cost per test
    $0.42
  36. 36
    Thinking MachinesDossier

    Accuracy

    Inkling, Accuracy: 0.5% (scale 0.0 to 100.0), 95% confidence interval 0.1 to 3.50.5%(0.1–3.5)
    Duration per test
    1,378.8 s
    Cost per test
    $2.50
  37. 37
    Thinking MachinesDossier

    Accuracy

    Inkling Small, Accuracy: 0% (scale 0.0 to 100.0), 95% confidence interval 0.0 to 0.00%(0.0–0.0)
    Duration per test
    705.1 s
    Cost per test
    $0.32
  38. 37
    NVIDIADossier

    high

    Accuracy

    Nemotron 3.5 Super VL, Accuracy: 0% (scale 0.0 to 100.0), 95% confidence interval 0.0 to 0.00%(0.0–0.0)
    Duration per test
    895.6 s
    Cost per test
    $0.86