MARINER - VQA Civilian Equipment Grounding: leaderboard

Metric: Accuracy (%) on MARINER maritime visual question answering (civilian equipment grounding questions; multiple-choice questions generated from verified image metadata and audited by maritime annotators); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 16 models tracked.

Top models

#ModelScore
1GPT-4.185.07
2Gemini 2.5 Pro84.18
3InternVL3-38B82.09
4InternVL3-78B81.49
5GPT-4o81.19
6InternVL3-8B77.01
7Qwen 2.5 VL 32B Instruct74.63
8Qwen 2.5 VL 7B Instruct73.73
9Qwen 2.5 VL 72B Instruct72.54
10Gemini 2.5 Flash68.96
11MiniCPM-V-2.668.06
12InternVL2-8B64.48

Interactive version: theaggregate.ai/benchmark?slug=mariner-vqa-civilian-equipment-grounding · How It Works · Data refreshed daily, snapshot 2026-10-07.