CursorBench 4.0 leaderboard
As of November 8, 2026, Claude Opus 5.5 (Anthropic) leads CursorBench 4.0 at 57.8%, ahead of Claude Sonnet 5.5 at 55.5% and Claude Fable 5.1 at 51.8%. 7 systems ranked. Every score here is published by the model's own lab; each row links to the page it comes from.
What is CursorBench 4.0? CursorBench is Cursor's internal coding benchmark, built from short and often ambiguous requests taken from real Cursor sessions; version 4.0 is the current one. Made by Cursor (Anysphere), September 2026.
Top models on CursorBench 4.0, November 2026
| # | Model | Score | Price / 1M | Context | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 Anthropic | adaptive thinking, max effort | $4 / $20 | 1M | Vendor-reported Anthropic |
| 2 | Claude Sonnet 5.5 Anthropic | best reported effort | $2 / $10 | 1M | Vendor-reported Anthropic |
| 3 | Claude Fable 5.1 Anthropic | as reported | $10 / $50 | 1M | Vendor-reported Anthropic |
| 4 | Claude Opus 5 Anthropic | as reported | $5 / $25 | 1M | Vendor-reported Anthropic |
| 5 | Grok 4.7 SpaceXSI | xhigh effort | $2 / $6 | 500K | Vendor-reported SpaceXSI (xAI) |
| 6 | Grok 4.6 SpaceXSI | high effort | $2 / $6 | 500K | Vendor-reported SpaceXSI (xAI) |
| 7 | Claude Sonnet 5 Anthropic | best reported effort | $2 / $10 | 1M | Vendor-reported Anthropic |
CursorBench 4.0: score against price
Can you trust CursorBench 4.0 scores?
No independent leaderboard publishes CursorBench 4.0 results for these models yet, so every number on this page is the lab's own report of its own model. Labs choose the harness, the effort level and the subset they report, and they rarely publish failures. Read the setting under each score before comparing two rows.
- The tasks are private and come mostly from Cursor's own code, so no one outside Cursor can reproduce or audit a score.
- Compare scores only within one version: the task mix changed across 3.0, 3.1, 3.2 and 4.0.
- Cursor itself says small differences may not be statistically meaningful.
How CursorBench 4.0 works
The task
Each task is a real request (an edit, a refactor, an investigation, following a design) that the agent carries out across several files, mostly in Cursor's own codebases.
The score
Model-based graders score each solution against the code that was actually committed; Cursor publishes a score per model and effort level with cost and tokens per task.
Test set
The authors do not publish a single task count for this version.
Made by
Cursor (Anysphere), first published September 2026.
Official sources
- cursor.com/en/cursorbench
- Definition read on 2026-11-08
Sources and freshness
7 published scores from 3 sources. Each was read on its source page; the oldest reading on this page is from November 8, 2026. Prices beside each score come from our model price index and move when a vendor reprices.
- Vendor-reported Anthropic
- Vendor-reported Anthropic
- Vendor-reported SpaceXSI (xAI)
CursorBench 4.0 FAQ
Answers derived from the 7 published scores on this page and the current price index.