LABBench2

Real-world capabilities of AI systems on scientific research tasks.

LABBench2 is a multi-task benchmark for measuring whether AI systems can perform the kinds of work scientists actually do — reading literature, interpreting figures and tables, reasoning about experimental protocols, and operating on biological sequences. It is composed of multiple sub-benchmarks, each isolating a specific skill. Results below are aggregated across the suite; click into a sub-benchmark for its individual leaderboard.

#	Model	Variant	Average score	Average output tokens	Coverage	Published
1	gpt-5-2	Tools + high	0.609	5.4k	14/15	2026-03-13
2	gpt-5-2-pro	Tools + high	0.583	3.7k	14/15	2026-03-13
3	gemini-3-pro-preview	Tools + high	0.564	1.3k	14/15	2026-03-13
4	claude-opus-4-6	Tools + high	0.519	6.9k	14/15	2026-03-23
5	claude-opus-4-5	Tools + high	0.471	3.6k	14/15	2026-03-23
6	gemini-3-pro-preview	Base	0.404	258	15/15	2026-03-13
7	gpt-5-2-pro	Base	0.382	1.7k	15/15	2026-03-13
8	claude-opus-4-6	Base	0.349	590	15/15	2026-03-23
9	gpt-5-2	Base	0.345	128	15/15	2026-03-13
10	claude-opus-4-5	Base	0.325	294	14/15	2026-03-23

Results

Leaderboard

Sub-benchmarks