# CursorBench 4.0 — AI model rankings

Cursor's own test of coding agents on ambiguous, multi-file tasks taken from real Cursor sessions — editing, refactoring, investigating a codebase, understanding what the user meant, managing jobs and following a design. Cursor runs each model at several reasoning efforts; each release here carries the score of its best listed effort. Scores aren't comparable with earlier CursorBench versions. Higher is better.

Scores from [CursorBench (Cursor)](https://cursor.com/cursorbench), which runs the benchmark and publishes the full field.

13 tracked models have a published CursorBench 4.0 score. Higher is better. Scores are gathered from CursorBench (Cursor); recorded retrieval dates appear beside the scores.

## Ranking

| Rank | Model | Developer | Score | Source | Released |
| --- | --- | --- | --- | --- | --- |
| 1 | Claude Opus 5.5 | Anthropic | 57.8% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Sep 22 2026 |
| 2 | Claude Sonnet 5.5 | Anthropic | 55.5% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Sep 28 2026 |
| 3 | Claude Fable 5.1 | Anthropic | 51.8% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Sep 1 2026 |
| 4 | Claude Opus 5 | Anthropic | 46.6% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Jul 24 2026 |
| 5 | Grok 4.7 | SpaceXAI | 46.3% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Sep 21 2026 |
| 6 | GPT-5.6 Sol | OpenAI | 41.7% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Jun 26 2026 |
| 7 | Muse Spark 1.3 | Meta | 41.6% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Sep 2 2026 |
| 8 | Grok 4.6 | SpaceXAI | 41.4% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Aug 12 2026 |
| 9 | GPT-5.6 Terra | OpenAI | 41.3% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Jun 26 2026 |
| 10 | Gemini 3.8 Flash | Google | 39.6% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Sep 2 2026 |
| 11 | GPT-5.6 Luna | OpenAI | 35.9% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Jun 26 2026 |
| 12 | Claude Sonnet 5 | Anthropic | 34.1% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Jun 30 2026 |
| 13 | Composer 2.5 | SpaceXAI | 27.7% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | May 18 2026 |


---

Canonical page: https://aireleasetracker.com/benchmark/cursorbench-4.0
Site index: https://aireleasetracker.com/llms.txt
Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.