CursorBench

Cursor leaderboard on ambiguous, multi-file coding tasks taken from real Cursor sessions, with cost, tokens and agent steps per task.

At a glance

Gemini 4 Neon Extra High leads with 65.4%, out of 53 results. Best score for the money: Haiku 5.5 Max, 48.4% for $1.14 per task. Cursor designs both the tasks and the agent that runs them: the ranking reflects work inside Cursor, and the cost per task shows what a higher reasoning level costs for a few more points.

Value and speed are picked among the top 27, so a cheap or quick run with a low score never wins. Data retrieved on Nov 22, 2026.

View on Cursor
Configurations
53
Best score
65.4%
Gemini 4 Neon Extra High
Lowest cost per task
$0.080
Haiku 5.5 Low · 27.8%
Version
4.1
Version Cursor names on its page

53 / 53 configurations

  1. #1
    Gemini 4 Neonxhigh

    Google

    65.4%

    Cost / task: $10.67Tokens / task: 136,273Steps / task: 73

  2. #2
    Gemini 4 Neonhigh

    Google

    63.7%

    Cost / task: $6.54Tokens / task: 88,596Steps / task: 56

  3. #3
    Gemini 4 Carbonhigh

    Google

    61.5%

    Cost / task: $8.52Tokens / task: 114,518Steps / task: 97

  4. #461.4%

    Cost / task: $8.09Tokens / task: 205,981Steps / task: 113

  5. #5
    Fable 5.2max

    Anthropic

    61.3%

    Cost / task: $48.11Tokens / task: 541,003Steps / task: 295

  6. #6
    Gemini 4 Carbonmedium

    Google

    60.4%

    Cost / task: $7.22Tokens / task: 103,189Steps / task: 83

  7. #7
    Fable 5.2xhigh

    Anthropic

    58.8%

    Cost / task: $17.84Tokens / task: 194,918Steps / task: 122

  8. #8
    Opus 5.5max

    Anthropic

    58.1%

    Cost / task: $11.88Tokens / task: 213,070Steps / task: 144

  9. #9
    Gemini 4 Neonmedium

    Google

    57.7%

    Cost / task: $3.07Tokens / task: 47,059Steps / task: 41

  10. #1057.3%

    Cost / task: $4.57Tokens / task: 113,807Steps / task: 77

  11. #11
    Opus 5.5xhigh

    Anthropic

    56.7%

    Cost / task: $6.54Tokens / task: 104,901Steps / task: 95

  12. #12
    Fable 5.2high

    Anthropic

    56.5%

    Cost / task: $10.51Tokens / task: 108,213Steps / task: 77

  13. #13
    Gemini 4 Argonhigh

    Google

    56.5%

    Cost / task: $9.14Tokens / task: 135,038Steps / task: 118

  14. #14
    Gemini 4 Argonmedium

    Google

    55.8%

    Cost / task: $7.64Tokens / task: 128,231Steps / task: 101

  15. #1555.0%

    Cost / task: $8.64Tokens / task: 48,526Steps / task: 127

  16. #16
    Opus 5.5high

    Anthropic

    54.5%

    Cost / task: $3.87Tokens / task: 56,162Steps / task: 64

  17. #17
    Sonnet 5.5max

    Anthropic

    54.4%

    Cost / task: $8.63Tokens / task: 342,273Steps / task: 199

  18. #18
    Fable 5.2medium

    Anthropic

    54.2%

    Cost / task: $5.01Tokens / task: 47,678Steps / task: 40

  19. #19
    Fable 5.2low

    Anthropic

    53.4%

    Cost / task: $4.26Tokens / task: 40,384Steps / task: 34

  20. #20
    Opus 5.5medium

    Anthropic

    50.0%

    Cost / task: $2.85Tokens / task: 39,736Steps / task: 52

  21. #21
    Sonnet 5.5xhigh

    Anthropic

    49.9%

    Cost / task: $3.13Tokens / task: 112,299Steps / task: 87

  22. #22
    Grok 4.7xhigh

    SpaceXSI

    48.6%

    Cost / task: $10.19Tokens / task: 90,268Steps / task: 110

  23. #23
    Haiku 5.5max

    Anthropic

    48.4%

    Cost / task: $1.14Tokens / task: 344,890Steps / task: 166

  24. #24
    Gemini 4 Argonlow

    —

    47.9%

    Cost / task: $7.96Tokens / task: 40,541Steps / task: 121

  25. #25
    Gemini 4 Hydrogenmedium

    Google

    46.4%

    Cost / task: $1.28Tokens / task: 43,606Steps / task: 45

  26. #26
    Grok 4.7high

    SpaceXSI

    46.0%

    Cost / task: $6.88Tokens / task: 65,850Steps / task: 84

  27. #27
    Gemini 4 Neonlow

    Google

    45.5%

    Cost / task: $1.26Tokens / task: 19,580Steps / task: 26

  28. #28
    Haiku 5.5xhigh

    Anthropic

    45.2%

    Cost / task: $0.56Tokens / task: 149,314Steps / task: 94

  29. #29
    Grok 4.6xhigh

    SpaceXSI

    44.3%

    Cost / task: $5.26Tokens / task: 58,554Steps / task: 62

  30. #30
    Sonnet 5.5high

    Anthropic

    43.8%

    Cost / task: $1.41Tokens / task: 43,088Steps / task: 48

  31. #31
    Opus 5.5low

    Anthropic

    42.7%

    Cost / task: $1.43Tokens / task: 19,497Steps / task: 32

  32. #32
    Grok 4.7medium

    —

    42.5%

    Cost / task: $5.52Tokens / task: 41,173Steps / task: 70

  33. #3341.9%

    Cost / task: $2.89Tokens / task: 56,572Steps / task: 113

  34. #34
    Grok 4.6high

    SpaceXSI

    41.3%

    Cost / task: $4.28Tokens / task: 46,330Steps / task: 52

  35. #35
    GPT-5.6 Solmax

    OpenAI

    40.9%

    Cost / task: $7.60Tokens / task: 40,466Steps / task: 100

  36. #36
    Grok 4.6medium

    SpaceXSI

    38.9%

    Cost / task: $2.59Tokens / task: 26,374Steps / task: 40

  37. #37
    Haiku 5.5high

    Anthropic

    37.5%

    Cost / task: $0.31Tokens / task: 77,196Steps / task: 61

  38. #38
    GPT-5.6 Solxhigh

    OpenAI

    37.4%

    Cost / task: $4.32Tokens / task: 23,921Steps / task: 58

  39. #3934.6%

    Cost / task: $2.33Tokens / task: 47,142Steps / task: 96

  40. #40
    Sonnet 5.5medium

    Anthropic

    34.3%

    Cost / task: $0.63Tokens / task: 19,144Steps / task: 26

  41. #41
    Haiku 5.5medium

    Anthropic

    34.2%

    Cost / task: $0.17Tokens / task: 43,441Steps / task: 41

  42. #42
    GPT-5.6 Solhigh

    OpenAI

    33.2%

    Cost / task: $3Tokens / task: 16,319Steps / task: 44

  43. #43
    Grok 4.6low

    SpaceXSI

    33.1%

    Cost / task: $1.76Tokens / task: 16,863Steps / task: 33

  44. #44
    Muse Spark 1.3high

    —

    32.1%

    Cost / task: $1.97Tokens / task: 36,562Steps / task: 85

  45. #45
    Sonnet 5.5low

    Anthropic

    31.8%

    Cost / task: $0.44Tokens / task: 13,798Steps / task: 21

  46. #46
    Grok 4.7low

    SpaceXSI

    31.5%

    Cost / task: $2.73Tokens / task: 14,493Steps / task: 42

  47. #47
    Muse Spark 1.3medium

    —

    31.3%

    Cost / task: $1.75Tokens / task: 32,034Steps / task: 77

  48. #48
    GPT-5.6 Solmedium

    OpenAI

    28.7%

    Cost / task: $1.97Tokens / task: 10,554Steps / task: 34

  49. #49
    Haiku 5.5low

    Anthropic

    27.8%

    Cost / task: $0.080Tokens / task: 23,673Steps / task: 27

  50. #50
    Muse Spark 1.3low

    —

    27.5%

    Cost / task: $1.13Tokens / task: 20,673Steps / task: 58

  51. #5123.9%

    Cost / task: $0.28Tokens / task: 10,497Steps / task: 22

  52. #52
    GPT-5.6 Sollow

    OpenAI

    23.0%

    Cost / task: $0.98Tokens / task: 5,206Steps / task: 22

  53. #53
    Muse Spark 1.3minimal

    —

    21.3%

    Cost / task: $0.69Tokens / task: 12,789Steps / task: 42

Method and versions

CursorBench is Cursor's in-house benchmark: agents solve ambiguous tasks spanning several files, drawn from real Cursor user sessions. BenchSift shows Cursor's public table as published, without recomputing it.

  • Tasks come from real Cursor sessions; Cursor does not publish the task set.
  • Each model is evaluated at one or more reasoning levels (Low to Max), each on its own row.
  • Cost, tokens and steps are per-task averages Cursor publishes.
  • A configuration is tied to a catalogue page by its model name and reasoning level; without a matching page it keeps Cursor's name.
  • Cursor designs both the benchmark and the agent that runs it: results reflect its own environment.

Data via Cursor, CursorBench 4.1.

Preferences and analytics

Rybbit: visits, performance, clicks, copies, forms and session replay (masked inputs), with your permission. Refusing does not limit the site.

Privacy

Allow navigation analytics?

With your permission, Rybbit measures visits, performance and interactions (clicks, copies, forms), including session replay, to improve BenchSift. Input values are masked in replays. Refusing does not limit the site.

Your choice is stored for 180 days. Change it at any time using “Privacy preferences” in the footer. Withdrawal reloads the page.

No choice stored: Rybbit is disabled.

Privacy