Benchmark family

GPQA

A family groups the versions of one benchmark. It is not evaluated and has no applicable models of its own; each of its 2 definitions is resolved separately.

General reasoningScience reasoningKnowledge

Definitions

Compare versions →

Newest first, following the recorded successor chain. Each definition is resolved against models separately.

BenchmarkGPQA
ActiveNewest in family
Scale
448 questions
Primary metric
Accuracy
Applicable models
13
BenchmarkGPQA Diamond
Active
Subset of GPQA
Scale
198 questions
Primary metric
Accuracy
Applicable models
13

Family methodology

No family-level methodology recorded.

Across definitions

Tasks: Graduate-level science multiple-choice Q&A
Sources: None recorded at family level
Last reviewed: 2026-09-22

Applicability does not mean evaluated; benchmark result/model score data is outside this UI contract.