FCMBench V1.1 - Document Type Recognition: leaderboard
Evaluates visual credit-document understanding on FCMBench V1.1: 26 document types, 5,198 images and 13,806 VQA pairs. Exact-match key/value sets are scored with F1; task scores macro-average fine-grained subtasks. Numeric calculation permits differences below 2. The paper prefers instruct or lowest available reasoning effort. The two-stage DeepSeek-OCR plus V3.2 system stays distinct from either model alone. These are paper V1.1 scores, not the differing repository leaderboard. This series reports FCMBench V1.1 - Document Type Recognition using F1 (%).
Metric: F1 (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 95.32 | #103 |
| 2 | Gemini 3 Pro | 90.68 | #82 |
| 3 | Claude Opus 4.5 | 84.07 | #87 |
| 4 | Kimi K2.5 | 79.07 | #152 |
| 5 | Qwen 3.5 397B A17B | 73.99 | #154 |
| 6 | GPT-5.1 | 72.29 | #145 |
| 7 | Qwen 3 VL 235B A22B Instruct | 66.65 | #280 |
| 8 | Qwen 3.5 35B A3B | 63.77 | #268 |
| 9 | GLM-4.6V | 62.89 | #330 |
| 10 | Qwen 3.5 122B A10B | 59.82 | #184 |
| 11 | GPT-5.2 | 59.2 | #115 |
| 12 | GLM-4.5V | 57.6 | #358 |
| 13 | Qwen 3 VL 32B Instruct | 54.28 | #290 |
| 14 | Qwen 3 VL 8B Instruct | 48.8 | #419 |
| 15 | Qwen 3 VL 30B A3B Instruct | 42.49 | #390 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=fcmbench-v1-1-document-type-recognition · How It Works · Data refreshed daily, snapshot 2026-11-05.