Coding · Benchmark profile
Cursor's first-party benchmark for ambiguous multi-file coding-agent tasks from real Cursor sessions (v3.2 public snapshot).
Data verified 15 July 2026 · Methodology 1.2.0
Benchmark score on CursorBench
Claude Fable 5 leads at 70.5%, followed by GPT-5.6 Sol (67.2%) and Grok 4.5 (66.7%).
Visual analysis
Three views of the same sourced leaderboard: model placement, the score curve, and descriptive provider averages. These visuals never alter ranking calculations.
10 ranked models · higher is better
| Model | Rank | Provider | Score |
|---|---|---|---|
| Claude Fable 5 | #1 | Anthropic | 70.5% |
| GPT-5.6 Sol | #2 | OpenAI | 67.2% |
| Grok 4.5 | #3 | xAI | 66.7% |
| GPT-5.6 Terra | #4 | OpenAI | 64.9% |
| Claude Opus 4.8 | #5 | Anthropic | 62.3% |
| Claude Sonnet 5 | #6 | Anthropic | 61.5% |
| GPT-5.6 Luna | #7 | OpenAI | 61.1% |
| GPT-5.5 | #8 | OpenAI | 58.4% |
| GLM-5.2 | #9 | Z.AI | 55% |
| Gemini 3.5 Flash | #10 | 48.8% |
Shows how quickly scores separate down the field
| Model | Rank | Provider | Score |
|---|---|---|---|
| Claude Fable 5 | #1 | Anthropic | 70.5% |
| GPT-5.6 Sol | #2 | OpenAI | 67.2% |
| Grok 4.5 | #3 | xAI | 66.7% |
| GPT-5.6 Terra | #4 | OpenAI | 64.9% |
| Claude Opus 4.8 | #5 | Anthropic | 62.3% |
| Claude Sonnet 5 | #6 | Anthropic | 61.5% |
| GPT-5.6 Luna | #7 | OpenAI | 61.1% |
| GPT-5.5 | #8 | OpenAI | 58.4% |
| GLM-5.2 | #9 | Z.AI | 55% |
| Gemini 3.5 Flash | #10 | 48.8% |
Average within the visible cohort · descriptive only
| Provider | Models | Average score |
|---|---|---|
| xAI | 1 | 66.7% |
| Anthropic | 3 | 64.8% |
| OpenAI | 4 | 62.9% |
| Z.AI | 1 | 55% |
| 1 | 48.8% |
One best score per model · higher is better
| Rank | Model | Provider | License | Score |
|---|---|---|---|---|
| #1 | Claude Fable 5 claude-fable-5 | Anthropic | closed | 70.5% |
| #2 | GPT-5.6 Sol gpt-5.6-sol | OpenAI | closed | 67.2% |
| #3 | Grok 4.5 grok-4.5 | xAI | closed | 66.7% |
| #4 | GPT-5.6 Terra gpt-5.6-terra | OpenAI | closed | 64.9% |
| #5 | Claude Opus 4.8 claude-opus-4-8 | Anthropic | closed | 62.3% |
| #6 | Claude Sonnet 5 claude-sonnet-5 | Anthropic | closed | 61.5% |
| #7 | GPT-5.6 Luna gpt-5.6-luna | OpenAI | closed | 61.1% |
| #8 | GPT-5.5 gpt-5.5 | OpenAI | closed | 58.4% |
| #9 | GLM-5.2 glm-5-2 | Z.AI | open | 55% |
| #10 | Gemini 3.5 Flash gemini-3.5-flash | closed | 48.8% |
The top of this snapshot is led by Claude Fable 5 at 70.5%; third place is 3.8 points behind. The top-10 spread is 21.7 points.
About CursorBench
Cursor's first-party benchmark for ambiguous multi-file coding-agent tasks from real Cursor sessions (v3.2 public snapshot). Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.
Open benchmark source ↗FAQ
Cursor's first-party benchmark for ambiguous multi-file coding-agent tasks from real Cursor sessions (v3.2 public snapshot).
Claude Fable 5 by Anthropic currently leads with 70.5%.
10 models in the LuminaBench cohort have a qualifying score on this benchmark.
Yes. This family is ranking-weighted at 3.2% of its capability category. Claude Fable 5 is overall #2.
Related
