SciVQR - Mathematics: leaderboard

The mathematics questions of SciVQR (2026), 3,254 image-based scientific questions from competitions, entrance exams and graduate-level sources in English and Chinese, answered with zero-shot chain-of-thought. Accuracy in percent; higher is better.

Metric: Mathematics accuracy (%) with zero-shot chain-of-thought prompting on SciVQR (3,254 image-based scientific questions from competitions, entrance exams and graduate-level sources, 78 percent multiple-choice and 22 percent free-form, in English and Chinese, high-school to graduate level); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1O4 Mini84.5#368
2Gemini 2.5 Flash (Preview 04-17)83.2#453
3O179.2#403
4MiMo-VL-7B-RL-250865.5#823
5Keye-VL-1.5-8BThinking61.1#1720
6Qwen 2.5 VL 72B Instruct54.3#841
7InternVL3-78B-Instruct47.7#1415
8GPT-4o47.4#751
9InternVL3-14B-Instruct36.9#1801
10InternVL3-8B-Instruct36.4#2872
11Qwen 2.5 VL 7B Instruct35.4#1902
12MiniCPM-o-2.626.8#2423
13LLaVA-OneVision-7B21.9#2981
14Molmo-7B-D-092420.3#7207
15LLaVA-NeXT-7B8#5032

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

Interactive version: theaggregate.ai/benchmark?slug=scivqr-mathematics · How It Works · Data refreshed daily, snapshot 2026-11-21.