DocFormBench (DocFormFlow): leaderboard
Metric: Format accuracy (%) on DocFormBench: share of required formatting attributes of the 500 content-aware Text-to-Format instances (Chinese and English .docx documents of 12 types) that exactly match the target, averaged per object type (page, paragraph, table, image) and then over types; the authors' DocFormFlow workflow (localize targets, then call formatting tools; temperature 1.0, up to 3 verification and self-repair rounds; multimodal models also receive page images, language models the text only); reasoning (thinking) disabled; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Minimal) | 80.36 |
| 2 | GLM-4.6 (Non-reasoning) | 75.72 |
| 3 | Qwen 3.5 397B A17B (Non-reasoning) | 75.19 |
| 4 | DeepSeek V4 Pro (Non-reasoning) | 74 |
| 5 | Qwen 3 VL 32B Instruct | 73.04 |
| 6 | DeepSeek V3.2 (Non-reasoning) | 72.87 |
| 7 | GPT-5 (Minimal) | 72.53 |
| 8 | Qwen 3 235B A22B 2507 Instruct | 71.18 |
| 9 | Qwen 3 VL 235B A22B Instruct | 70.94 |
| 10 | Qwen 3.5 122B A10B (Non-reasoning) | 70.87 |
| 11 | GLM-4.6V (Non-reasoning) | 65.78 |
| 12 | Qwen 3 VL 30B A3B Instruct | 64.47 |
| 13 | DeepSeek V4 Flash (Non-reasoning) | 64.36 |
| 14 | Qwen 3.5 35B A3B (Non-reasoning) | 62.1 |
| 15 | Qwen 3 VL 8B Instruct | 47.64 |
Interactive version: theaggregate.ai/benchmark?slug=docformbench-docformflow · How It Works · Data refreshed daily, snapshot 2026-09-29.