GPQA A
AI benchmark · Knowledge
What is GPQA?
GPQA measures model accuracy on 448 extremely difficult, Google-proof graduate-level multiple-choice questions in biology, physics, and chemistry.
Science mostly
“448 multiple-choice questions written by domain experts in biology, physics, and chemistry”
arxiv.org/abs/2311.12022 ↗“448 multiple-choice questions written by domain experts in biology, physics, and chemistry”
arxiv.org/abs/2311.12022 ↗
Paper: arxiv.org/abs/2311.12022 ↗
Lineage
Benchmarks this one is built on and built from: versions, subsets, extensions and the indexes that combine them.
Derived benchmarks
GPQA Diamond subset
Included in
“MMLU-Pro, GPQA, AIME, HLE, ARC-AGI-2, Metacog, SWE-Pro, IFEval, LCB, **Union Eval**, etc.”
huggingface.co/datasets/youssef3146/ALL-Bench-Leaderboard ↗“Open LLM Leaderboard v2 compares open large language models in a reproducible way, scoring them on IFEval, BBH, MATH, GPQA, MuSR and MMLU-PRO.”
Tier A
Widely used: a lab reports it, or 50 or more models are scored on it.
How much stands behind a tier: usage, who backs it, a confirmed line from its own site or paper, and a clear read on X. High with three of the four, medium with two, low otherwise.
- no usage data
- in curated RL directories (1 pt)
- its family includes GPQA Diamond (tier S)
On X · mixed, moderate buzz, from 14 posts about GPQA in the last month
GPQA Diamond remains a standard line in model release and quantization tables, but scores near 90% and sarcasm about saturation suggest it is losing discriminative power.
How we got to tier A
- Adoption: no usage data. Tier D.
- On X: mixed, moderate buzz. “GPQA Diamond remains a standard line in model release and quantization tables, but scores near 90% and sarcasm about saturation suggest it is losing discriminative power.” No move.
- In curated RL directories: 1 point, a floor of none.
- Root of a benchmark family that includes GPQA Diamond (tier S): raised to A, one step below.
- Confirmed: a line from its own site or paper places it.
A clear view on X moves a tier one step up or down; most entries don't move. How tiers work.
Where GPQA sits
Its domain on the map, live. Hover a logo for its name; click to open it. Open it on the full map
Next: browse all benchmarks. Entry from the RL Research daily scrape of public sources, 2026-11-05 snapshot.