CrossVid - Comparative Culinary QA: leaderboard

Open-ended answers comparing cooking videos, scored by GPT-4.1 for coverage and correctness of the reference key points; unlike the other CrossVid boards, no answer options are given.

Metric: Judge score (%). Source: raw.githubusercontent.com. Saturation forecast: Estimated already saturated. 15 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Pro59.8#274
2Qwen 2.5 VL 32B Instruct41.2#1074
3Kimi-VL-A3B (Thinking)29.2
4GLM-4.1V-9B (Thinking)26.9
5InternVL3-78B23.2#755
6LLaVA-Video-72B-Qwen217.8#1344
7InternVL3-38B16.2#927
8Qwen2.5-Omni-7B15.3#1883
9Qwen 2.5 VL 7B Instruct12#1893
10LongVA-7B-DPO10.7#2177
11VideoLLaMA3-7B9.8#2015
12InternVL3-8B9.7#1756
13MiniCPM-o-2.69#2417
14Video-R1-7B8#2053
15Phi-3.5-Vision Instruct4.3#6271

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

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