MARINER - VQA OCR-Integrated QA: leaderboard
Metric: Accuracy (%) on MARINER maritime visual question answering (ship text questions; multiple-choice questions generated from verified image metadata and audited by maritime annotators); higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | MiniCPM-V-2.6 | 91.58 |
| 2 | InternVL3-78B | 90.1 |
| 3 | InternVL3-38B | 89.85 |
| 4 | Qwen 2.5 VL 32B Instruct | 89.11 |
| 5 | Qwen 2.5 VL 72B Instruct | 88.86 |
| 6 | InternVL2-8B | 88.61 |
| 7 | Qwen 2.5 VL 7B Instruct | 88.61 |
| 8 | InternVL3-8B | 87.38 |
| 9 | Gemini 2.5 Pro | 86.39 |
| 10 | GPT-4.1 | 84.16 |
| 11 | GPT-4o | 78.22 |
| 12 | Gemini 2.5 Flash | 58.42 |
Interactive version: theaggregate.ai/benchmark?slug=mariner-vqa-ocr-integrated-qa · How It Works · Data refreshed daily, snapshot 2026-10-07.