Verified AI benchmark

GPQA Diamond Leaderboard

Publicly reported reasoning benchmark result.

Semantic classTechnical Performance CategoryReasoning DirectionHigher is better Verified entities23

How to interpret this leaderboard

What this benchmark does—and does not—measure

This is a technical-performance benchmark. Compare scores only within this benchmark definition and version; AI Orbit does not combine it with product-experience or otherwise incompatible semantic classes.

Method & source

Unit: %
Results: 23
Latest test: Aug 7, 2025
Distinct source links: 9

Open official methodology / source

Leading score96.00 %

Entities at top score1

Evidence freshnessAug 7, 2025

Verified results

GPQA Diamond rankings

Only the latest verified result for each listed entity is shown. Equal scores receive the same displayed rank.

Rank	Model / Tool	Score	Tested	Evidence
#1	GPT-6 Astra	96.00%	Not recorded	OpenAI
#2	GPT-5.6 Sol	94.60%	Not recorded	OpenAI
#2	GPT-5.6	94.60%	Not recorded	OpenAI
#4	Claude Fable 5.1	93.70%	Not recorded	OpenAI cross-provider evaluation
#5	GPT-5.6 Terra	92.90%	Not recorded	OpenAI
#6	GPT-5.6 Luna	92.30%	Not recorded	OpenAI
#7	Gemini 2.5 Pro	86.40%	Not recorded	Google DeepMind
#8	GPT-5	85.70%	Aug 7, 2025	OpenAI GPT-5 developer announcement
#9	o3	83.30%	Aug 7, 2025	OpenAI GPT-5 developer announcement
#10	Gemini 2.5 Flash	82.80%	Not recorded	Google DeepMind
#11	GPT-5 mini	82.30%	Aug 7, 2025	OpenAI GPT-5 developer announcement
#12	o4-mini	81.40%	Aug 7, 2025	OpenAI GPT-5 developer announcement
#13	Claude Opus 4	79.60%	May 22, 2025	Anthropic Claude 4 announcement
#14	Claude 3.7 Sonnet	78.20%	May 22, 2025	Anthropic Claude 4 announcement
#15	Claude Sonnet 4	75.40%	May 22, 2025	Anthropic Claude 4 announcement
#16	DeepSeek-R1	71.50%	Not recorded	DeepSeek
#17	GPT-5 nano	71.20%	Aug 7, 2025	OpenAI GPT-5 developer announcement
#18	GPT-4.1	66.30%	Aug 7, 2025	OpenAI GPT-5 developer announcement
#19	GPT-4.1 mini	65.00%	Aug 7, 2025	OpenAI GPT-5 developer announcement
#20	GPT-4.1 nano	50.30%	Aug 7, 2025	OpenAI GPT-5 developer announcement
#21	GPT-4o	46.00%	Apr 14, 2025	OpenAI GPT-4.1 announcement
#22	Gemma 3 27B	42.40%	Not recorded	Google
#23	GPT-4o mini	40.20%	Apr 14, 2025	OpenAI GPT-4.1 announcement

Limitations

What to keep in mind

A benchmark is one evidence signal, not a universal product verdict. Test conditions, benchmark versions, source methodology and date can all affect interpretation. Compare shared benchmarks and current product evidence before making a model or tool decision.

Freshness

Why dates and sources are shown

AI Orbit retains the source and tested date for each verified result so readers can judge whether evidence is current enough for their use case. Older evidence is not silently presented as a new test.