CiteVQA: leaderboard
Multimodal Large Language Models (MLLMs) have significantly advanced document understanding, yet current Doc-VQA evaluations score only the final answer and leave the supporting evidence unchecked. This answer-only approach masks a critical failure mode: a model can land on the correct answer while grounding it in the wrong passage -- a critical risk in high-stakes domains like law, finance, and m
Metric: Overall SAA (self-reported). Source: benchmarklist.com. Saturation forecast: Around May 2027. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 76 | #58 |
| 2 | Gemini 3 Flash (Preview) | 65.4 | #82 |
| 3 | GPT-5.4 | 59 | #79 |
| 4 | Gemini 2.5 Pro | 47 | #152 |
| 5 | Seed 2.0 Pro | 44.1 | #118 |
| 6 | GPT-5.2 | 33.7 | #110 |
| 7 | Qwen 3 VL 235B A22B | 22.5 | #238 |
| 8 | Gemma 4 31B | 20.2 | #200 |
| 9 | Kimi K2.5 | 19.1 | #144 |
| 10 | Qwen 3.5 397B A17B | 18.3 | #146 |
| 11 | Qwen 3.6 Plus | 17.5 | #120 |
| 12 | Qwen 3.5 27B | 17.3 | #187 |
| 13 | Qwen 3.5 122B A10B | 14.8 | #173 |
| 14 | GLM-5V Turbo | 12.8 | #166 |
| 15 | Qwen 3.5 9B | 11.1 | #379 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=citevqa · How It Works · Data refreshed daily, snapshot 2026-10-20.