Real5-OmniDocBench - Skew: leaderboard
Metric: Overall (%) in the Skew scenario: the OmniDocBench v1.5 overall score ((1 - text normalized edit distance) x 100 + formula CDM + table TEDS, divided by 3) for end-to-end document parsing of the 1,355 OmniDocBench v1.5 pages printed and physically recaptured, photographed at pitch, roll, yaw, compound and extreme perspective tilts; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 89.45 | #77 |
| 2 | Gemini 2.5 Pro | 89.07 | #145 |
| 3 | Kimi K2.5 | 88.86 | #139 |
| 4 | Qwen 2.5 VL 72B Instruct | 86.9 | #364 |
| 5 | Qwen 3 VL 235B A22B | 86.56 | #228 |
| 6 | GPT-5.2 | 75 | #105 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=real5-omnidocbench-skew · How It Works · Data refreshed daily, snapshot 2026-10-11.