GB/T-Bench - Diagnosis Recall - Document Structure: leaderboard
Metric: Diagnosis recall (%; exact section, dimension and error-type match). Source: arxiv.org. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.6 Sol | 75.81 |
| 2 | GPT-5.5 | 71.94 |
| 3 | Claude Fable 5 | 65.08 |
| 4 | GPT-5.4 | 55.87 |
| 5 | Qwen 3.7 Plus | 54.41 |
| 6 | DeepSeek V4 Pro | 50.33 |
| 7 | Gemini 3.5 Flash | 49.87 |
| 8 | GPT-5.3 | 48.26 |
| 9 | DeepSeek V4 Flash | 41.81 |
| 10 | Claude Opus 4.8 | 35.57 |
| 11 | GPT-5.4 Mini | 31.99 |
| 12 | MiniMax-M2.7 | 29.22 |
| 13 | Qwen 2.5 VL 72B Instruct | 25.34 |
| 14 | Qwen 3 235B A22B 2507 Instruct | 24.74 |
Interactive version: theaggregate.ai/benchmark?slug=gb-t-bench-diagnosis-recall-document-structure · How It Works · Data refreshed daily, snapshot 2026-09-19.