Virology Capabilities Test: leaderboard

322 multimodal virology questions with expert percentiles. o3 scores 43.8% (94th percentile among PhD virologists who average 22.1%). Data suggests LLMs may surpass all human experts within 4-5 years.

Metric: Accuracy (%). Source: www.virologytest.ai. 8 models tracked.

Top models

#ModelScore
1O343.8
2Gemini 2.5 Pro37.6
3O4 Mini37
4O135.4
5Claude 3.7 Sonnet30.8
6GPT-4.5 (Preview)28.3
7Median Expert22.1
8GPT-4o18.8

Interactive version: theaggregate.ai/benchmark?slug=virology-capabilities-test · How It Works · Data refreshed daily, snapshot 2026-09-05.