Sci-Rho - Biology (English): leaderboard

Metric: Average-case accuracy (%) on the English biology templates of Sci-Rho: 606 expert-written executable templates per language (mathematics, physics, chemistry, biology and computer science, many drawn from olympiad problems), each rendered as 10 visually and numerically varied image-grounded instances; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 17 models tracked.

Top models

#ModelScore
1GPT-5.4 (Non-reasoning)72.2
2Gemini 2.5 Pro71.7
3Qwen 3.5 122B A10B (Non-reasoning)67.8
4Qwen 3.5 27B (Non-reasoning)65.7
5Qwen 3 VL 8B Instruct53.5
6Gemini 2.5 Flash (Non-reasoning)51.7
7Llama 4 Scout Instruct44.3
8InternVL3.5-8B36.5
9Gemma 3 12B (IT)34.3
10Molmo2-8B26.1
11Qwen 3 VL 4B Instruct17.8
12Gemma 3 4B (IT)16.1

Interactive version: theaggregate.ai/benchmark?slug=sci-rho-biology-english · How It Works · Data refreshed daily, snapshot 2026-09-29.