MARINER - VQA OCR-Integrated QA: leaderboard

Maritime visual question answering on MARINER (2026), with multiple-choice questions built from verified image metadata and audited by maritime annotators. Accuracy in percent on questions that need text read from the ship; higher is better.

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. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1MiniCPM-V-2.691.58#3311
2InternVL3-78B90.1#765
3InternVL3-38B89.85#948
4LLaVA-OneVision-72B89.85#1782
5Qwen 2.5 VL 32B Instruct89.11#1065
6Qwen 2.5 VL 72B Instruct88.86#849
7Qwen 2.5 VL 7B Instruct88.61#1919
8InternVL2-8B88.61#3268
9InternVL3-8B87.38#1795
10Gemini 2.5 Pro86.39#280
11Qwen 2.5 VL 3B Instruct85.89#2720
12GPT-4.184.16#506
13LLaVA-OneVision-7B83.17#3002
14GPT-4o78.22#754
15LLaVA-v1.5-7B66.58#7642

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=mariner-vqa-ocr-integrated-qa · How It Works · Data refreshed daily, snapshot 2026-11-22.