Skip to main content
idapt
AccueilModèles IATarifs
Sign inStart free
All terms

Learn

What is GPQA?

Last reviewed July 16, 2026

GPQA (Graduate-Level Google-Proof Q&A) is a benchmark of PhD-level multiple-choice questions in biology, physics, and chemistry, written and validated by domain experts. The commonly reported GPQA Diamond subset has 198 questions chosen to be search-resistant: experts score around 65%, while skilled non-experts with web access score little better than chance. It is the standard measure of expert scientific reasoning in AI models.

Why Google-proof matters

Questions were kept only when experts answered them correctly and skilled non-experts failed even with the web available. That construction filters out lookup and pattern-matching: a high GPQA score means the model reasons through graduate-level material rather than retrieving it. It also makes the benchmark harder to game with training-data contamination.

How to read a GPQA score

Scores are the fraction of the 198 Diamond questions answered correctly. The expert human baseline is about 65%, so models above it exceed domain experts on this format. Each question is worth about half a percentage point, so treat close scores as ties, and weigh GPQA alongside broader suites.

Top models on GPQA

Full leaderboard
  1. 1GPT 5.4 Pro94.6%
  2. 2GPT 5.6 Sol94.6%
  3. 3Gemini 3.1 Pro94.1%
  4. 4GPT 5.594.0%
  5. 5GPT 5.5 Pro93.9%
  6. 6Kimi K393.5%
  7. 7Grok 4.593.4%
  8. 8GPT 5.493.3%
  9. 9GPT 5.6 Terra92.9%
  10. 10Gemini 3.5 Flash92.8%

Peak scores from the compiled benchmark data; reasoning models use their highest effort tier.

In idapt

idapt's rankings order the catalog by GPQA Diamond, model pages show each model's score beside its price, and the compare view puts several models' benchmark rows side by side.

GPQA leaderboardCompare models

Frequently asked

Related terms

Humanity's Last ExamMMLU-ProEpoch Capabilities Index

Multi-Model Chat

GPT, Claude, Gemini, Grok, Llama: pick mid-conversation, cheapest-first.

Agents

Skills, scoped permissions, persistent memory: any model you pick.

See it in practice

The fastest way to understand it is to use it. Start free, no credit card.

Try idapt free
  • Accueil
  • Tarifs
  • Modèles IA
  • Modèles d'image
  • Modèles de voix
  • Modèles vidéo
  • Classements
  • New models
  • Model status
  • Multi-Model Chat
  • Voice Mode
  • Voice HUD
  • Web Search
  • Image Generation
  • Video Generation
  • Audio Generation
  • Transcription
  • Drive
  • Secrets
  • Sharing
  • Workspaces
  • Tasks
  • Memory
  • Agents
  • Subagents
  • Automations
  • Skills
  • Code Execution
  • Computers
  • Computer Use
  • Cloud Computers
  • Local AI
  • AI Gateway
  • API & SDK
  • CLI
  • MCP
  • Tunnels
  • All features →
  • Blog
  • Cas d'usage
  • Comparisons
  • Best of
  • Learn
  • Free tools
  • Prompts
  • Templates
  • Changelog
  • Centre d'aide
  • FAQ
  • Confidentialité
  • Comparer tous les modèles
  • Support
  • Développeurs
  • Quickstarts
  • Référence API
  • Tarifs API
  • CLI
  • MCP
  • Downloads
  • Desktop
  • Badges and embeds
© idapt[email protected]ConditionsPolitique de confidentialitéMentions légalesSignaler un contenu
X (Twitter)