Model results

GPQA Diamond

Choose a shared evaluation protocol to compare model scores. Versions, sources and configurations stay separate.

Atlas snapshot captured 10/19/2026, 10:20:28 AM UTC. This is an import timestamp, not the evaluation date. Coverage may be incomplete.

Reviewed recommendation and limits

Use as a compact reference for expert-written science questions, alongside evidence that matches the scientific work you intend to do.

Original GPQA Diamond subset. Preserve prompt, answer extraction, tools, sampling and aggregation. Keep main, extended and third-party corrected subsets separate.. Imported groups below may use different protocols; this review does not certify each run. The static 198-question subset has limited statistical power and exposure risk. Its expert filtering does not make it representative of scientific practice. A Google-proof collection process does not prove present-day contamination resistance.

Publisher results GPQA paper

← All evaluation groups

gpqa · diamond

Source: Hugging Face eval result
Version: Not reported
Measurement date: 2026-01-27
Metric: % · Higher is better
Evaluation scope: Registered model evaluation · Model Card

Positions are withheld because at least one result lacks explicit independent provenance or a documented evaluation scope. Missing metadata is shown as unreported; matching imported metadata is not independent verification of the protocol.

Kimi K2 Thinking
moonshotai/kimi-k2-thinking-20251106
84.5%
View modelEvidence Evidence class unreported
GLM 4.7 Flash
z-ai/glm-4.7-flash-20260119
75.2%
View modelEvidence Evidence class unreported
R1
deepseek/deepseek-r1
71.5%
View modelEvidence Evidence class unreported
Llama 3.1 8B Instruct
meta/llama-3.1-8b-instruct
30.4%
View modelEvidence Evidence class unreported