Benchmark profile

CursorBench

Cursor's current first-party benchmark for ambiguous, multi-file coding-agent tasks from real Cursor sessions.

Data verified July 23, 2026

Benchmark score on CursorBench — July 23, 2026

BenchLM mirrors the published score view for CursorBench. Claude Fable 5 leads the public snapshot at 70.5% , followed by GPT-5.6 Sol (67.2%) and Grok 4.5 (66.7%). BenchLM does not use these results to rank models overall.

1Closed

Claude Fable 5

Anthropic

claude-fable-5

70.5%

Overall 83.68Context 1M+

2Closed

GPT-5.6 Sol

OpenAI

gpt-5-6-sol

67.2%

Overall 81.96Context 1M

3Closed

Grok 4.5

xAI

grok-4-5

66.7%

Overall 76.72Context 500K

13 modelsCodingRefreshingDisplay onlyUpdated July 23, 2026

Benchmark score table (13 models)

Score

Claude Fable 5Anthropic · Closed

70.5%

GPT-5.6 SolOpenAI · Closed

67.2%

Grok 4.5xAI · Closed

66.7%

GPT-5.6 TerraOpenAI · Closed

64.9%

Claude Opus 4.8Anthropic · Closed

62.3%

Claude Sonnet 5Anthropic · Closed

61.5%

GPT-5.6 LunaOpenAI · Closed

61.1%

GPT-5.5OpenAI · Closed

58.4%

Composer 2.5Cursor · Closed

56.1%

GLM-5.2Z.AI · Open weight

55.0%

Gemini 3.6 FlashGoogle · Closed

53.5%

Kimi K2.7 CodeMoonshot AI · Open weight

49.7%

Gemini 3.5 FlashGoogle · Closed

48.8%

The published CursorBench snapshot places Claude Fable 5 first at 70.5%. The third row is 3.8 points behind. The broader top-10 range is 15.5 points, so the table still separates the published systems.

13 models have been evaluated on CursorBench. The benchmark falls in the Coding category. This category carries a 20% weight in BenchLM.ai's overall scoring system. CursorBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About CursorBench

Year

2026

Tasks

Harder long-horizon agentic coding tasks

Format

Cursor agent-loop evaluation

Difficulty

Professional agentic software engineering

Cursor reports CursorBench v3.2 on its public evals page for ambiguous, multi-file tasks from real Cursor sessions. BenchLM tracks it as display-only because it is a first-party benchmark and flags Cursor's published Grok 4.5 contamination caveat in provenance.

CursorBench 3.2

BenchLM freshness & provenance

Version

cursorBench32

Refresh cadence

Static

Staleness state

Refreshing

Question availability

Public benchmark set

RefreshingDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

FAQ

What does CursorBench measure?

Cursor's current first-party benchmark for ambiguous, multi-file coding-agent tasks from real Cursor sessions.

Which model scores highest on CursorBench?

Claude Fable 5 by Anthropic currently leads with a score of 70.5% on CursorBench.

How many models are evaluated on CursorBench?

13 AI models have been evaluated on CursorBench on BenchLM.

Compare Top Models on CursorBench

Claude Fable 5 vs GPT-5.6 Sol GPT-5.6 Sol vs Grok 4.5 Grok 4.5 vs GPT-5.6 Terra GPT-5.6 Terra vs Claude Opus 4.8

Last updated: July 23, 2026 · BenchLM version cursorBench32

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.