Frontier baseline
GPQA Diamond
Graduate-level science questions designed to be hard for non-experts and still hard for models.
As of 13 August 2026. Conditional pack. GPQA Diamond. Component of the AA Intelligence Index. No Fable 5 or GPT-5.6 Diamond cell is copied without a published number.
GPQA Diamond is graduate-level science QA. Official accuracy is the load-bearing metric. Fallback to another model is a different outcome.
Artificial Analysis recorded Fable 5 fallback on scientific items, including GPQA, as part of Intelligence Index work. That is not a Diamond percentage. Recheck terms before a commercial run.
- Scale
- GPQA Diamond split. Fable 5 can fall back to Opus 4.8 on scientific items in the AA harness.
- Provenance
- Rein et al., GPQA. Component of Artificial Analysis Intelligence Index v4.1.1. Recheck terms before a commercial run.
- Contamination
- High. Public since 2023.
- Cannot establish
- Professional-document workflow performance. No Fable 5 or GPT-5.6 Diamond cell is copied here without a published number.
Metrics, kept separate
- Official accuracy
- Fallback or refusal rate
No cited public figure is installed for this pack. Method, scale, and gates still apply. A missing chart is not a pass rate.
What the figures bound
The pack is listed because knowledge-work buyers still cite GPQA as a frontier baseline. It is not filing QA and not a diligence memo score.
No chart is installed. HLE and the Intelligence Index carry the dated Fable 5 cells that AA did publish as numbers.
What this page does not conclude
- Professional-document workflow performance
- A copied Fable 5 or GPT-5.6 Diamond percentage
- A commercial run before terms are approved
- A science-exam winner used as a product score
Cited as calibration on these memos — not a score of those products.
Not attached to a software, advisory, or automation memo yet.