# GPQA

> GPQA (Graduate-Level Google-Proof Q&A) is a benchmark of difficult multiple-choice questions in biology, physics, and chemistry, written by domain experts so that skilled non-experts cannot answer them reliably even with unrestricted web access.

- Identifier: PTL-0189
- Category: Benchmarks
- Canonical URL: https://protologue.com/t/gpqa/
- Also known as: GPQA Diamond, Graduate-Level Google-Proof Q&A
- Introduced: 2023

## Description

Its high-quality "Diamond" subset is the version most often reported in model announcements. Because the questions resist search, it measures expert-level reasoning more than retrieval. It is also used to study scalable oversight, the question of how people can supervise models that know more than they do.

## Broader terms

- [Benchmark](https://protologue.com/t/benchmark/)

## Related terms

- [MMLU-Pro](https://protologue.com/t/mmlu-pro/)
- [Humanity's Last Exam](https://protologue.com/t/humanitys-last-exam/)
- [Reasoning Model](https://protologue.com/t/reasoning-model/)

## Sources

- Rein et al. (2023). GPQA: A Graduate-Level Google-Proof Q&A Benchmark. https://arxiv.org/abs/2311.12022

## Cite this entry

Protologue. (2026). GPQA. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0189). https://protologue.com/t/gpqa/

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)