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: Epoch AI Benchmarking Hub
Version: Not reported
Measurement date: Not reported
Metric: % · Higher is better
Evaluation scope: Epoch AI run · task v1 · effort max · result date not supplied

Exact score ties share a position. Positions do not establish statistical significance. Missing metadata is shown as unreported; matching imported metadata is not independent verification of the protocol.

#1GPT-6 Astra
openai/gpt-6-astra-20260903
95.8%
View modelEvidence independent
#2Claude Sonnet 5.5
anthropic/claude-sonnet-5-5
95.6%
View modelEvidence independent
#3GPT-6.1 Sol
openai/gpt-6.1-sol-20260929
95.4%
View modelEvidence independent
#4GPT-6 Sol
openai/gpt-6-sol-20260922
94.3%
View modelEvidence independent
#5Claude Opus 5
anthropic/claude-opus-5
93.9%
View modelEvidence independent
#6GPT-5.6 Sol
openai/gpt-5.6-sol-20260709
93.5%
View modelEvidence independent
#7GPT-5.6 Terra
openai/gpt-5.6-terra-20260709
93.3%
View modelEvidence independent
#8Kimi K3
moonshotai/kimi-k3-20260715
93.1%
View modelEvidence independent
#9GLM 5.2
z-ai/glm-5.2-20260616
91.9%
View modelEvidence independent
#10DeepSeek V4 Pro 0813
deepseek/deepseek-v4-pro-20260813
91.7%
View modelEvidence independent
#11GPT-5.6 Luna
openai/gpt-5.6-luna-20260709
91.6%
View modelEvidence independent
#12Claude Opus 4.8
anthropic/claude-opus-4-8
91%
View modelEvidence independent
#12DeepSeek V4 Flash 0731
deepseek/deepseek-v4-flash-20260731
91%
View modelEvidence independent
#14GLM 5.3
z-ai/glm-5.3-20260816
90.9%
View modelEvidence independent
#15Claude Opus 5.5
anthropic/claude-opus-5-5
90.6%
View modelEvidence independent
#16GPT-6 Luna
openai/gpt-6-luna-20260922
90.5%
View modelEvidence independent
#17GLM 5.3 Flash
z-ai/glm-5.3-flash-20260826
90.2%
View modelEvidence independent
#18DeepSeek V4.1 Flash
deepseek/deepseek-v4.1-flash-20260910
89.8%
View modelEvidence independent
#19Claude Haiku 5.5
anthropic/claude-haiku-5-5
89.6%
View modelEvidence independent
#19DeepSeek V4 Pro 0423
deepseek/deepseek-v4-pro-20260423
89.6%
View modelEvidence independent
#21Claude Opus 4.6
anthropic/claude-opus-4-6
88.4%
View modelEvidence independent
#22Claude Opus 4.7
anthropic/claude-opus-4-7
86.4%
View modelEvidence independent
#23Claude Fable 5
anthropic/claude-fable-5
85.9%
View modelEvidence independent
#24Claude Sonnet 5
anthropic/claude-sonnet-5
80.3%
View modelEvidence independent
#25Claude Sonnet 4.6
anthropic/claude-sonnet-4-6
78.8%
View modelEvidence independent