MARINER - Vessel Classification (Macro-F1): leaderboard

Fine-grained vessel classification on MARINER (2026), a maritime vision benchmark of shore, shipboard, drone and open-source images with 63 vessel categories and adverse weather. Macro-F1 over vessel categories in percent; higher is better.

Metric: Macro-F1 (%) over vessel categories on MARINER fine-grained vessel classification (63 categories: 35 military, 27 civilian and Other; maritime images from shore, shipboard, UAV and open sources, including fog, ice, night and wave conditions); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-4.184.14#505
2GPT-4o80.89#751
3Gemini 2.5 Flash80.36#463
4Qwen 2.5 VL 72B Instruct63.06#841
5Gemini 2.5 Pro59.68#276
6Qwen 2.5 VL 32B Instruct59.22#1062
7InternVL3-78B52.78#759
8Qwen 2.5 VL 7B Instruct51.18#1902
9InternVL3-38B49.37#939
10LLaVA-OneVision-72B47.12#1771
11Qwen 2.5 VL 3B Instruct37.16#2713
12MiniCPM-V-2.635.92#3267
13LLaVA-OneVision-7B28.14#2981
14InternVL3-8B22.03#1782
15InternVL2-8B13.67#3227

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=mariner-vessel-classification-macro-f1 · How It Works · Data refreshed daily, snapshot 2026-11-21.