CrossVid - Multi-View Reasoning: leaderboard

Mean accuracy over CrossVid's two multi-view tasks: spatial reasoning and object counting across drone videos of one scene from different viewpoints.

Metric: Accuracy (%). Source: raw.githubusercontent.com. Saturation forecast: Estimated already saturated. 16 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1InternVL3-8B40.7#1756
2InternVL3-38B38.6#927
3GPT-4.138.4#505
4GLM-4.1V-9B (Thinking)37.8
5Kimi-VL-A3B (Thinking)32.7
6Qwen 2.5 VL 32B Instruct31.7#1074
7MiniCPM-o-2.631.4#2417
8Qwen2.5-Omni-7B29.6#1883
9Gemini 2.5 Pro28.7#274
10InternVL3-78B28.1#755
11LLaVA-Video-72B-Qwen227.9#1344
12Phi-3.5-Vision Instruct27.6#6271
13Video-R1-7B26.9#2053
14LongVA-7B-DPO26.3#2177
15VideoLLaMA3-7B19.8#2015

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=crossvid-multi-view-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-17.