MEGA-Bench Task - Doc VQA: leaderboard
The "Doc VQA" subtask of MEGA-Bench Task. A single task from MEGA-Bench, a large-scale multimodal evaluation suite that scores vision-language models across 500+ real-world tasks spanning perception, reasoning, OCR, charts, documents, and multi-image understanding.
Metric: Task Score (%). Source: huggingface.co. Saturation forecast: Around January 2027. 44 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20241022) | 92.5 | #291 |
| 2 | Claude 3.5 Sonnet (20240620) | 87.5 | #299 |
| 3 | GPT-4o Mini | 83.8 | #599 |
| 4 | Gemini 2.5 Pro (Preview 03-25) | 83.8 | #123 |
| 5 | Pixtral-12B | 83.1 | #808 |
| 6 | GPT-4o | 82.5 | #343 |
| 7 | Gemini 1.5 Flash (002) | 81.9 | #583 |
| 8 | Llama 4 Scout Base | 81.9 | #707 |
| 9 | InternVL3-8B | 80.6 | #620 |
| 10 | Gemini 1.5 Pro (002) | 80.6 | #349 |
| 11 | InternVL3-14B | 80.6 | #505 |
| 12 | Qwen 2 VL 72B | 80 | #460 |
| 13 | InternVL2-8B | 78.8 | #817 |
| 14 | Gemma 3 12B (IT) | 76.9 | #664 |
| 15 | InternVL3-78B | 76.9 | #352 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mega-bench-task-doc-vqa · How It Works · Data refreshed daily, snapshot 2026-10-20.