GPT-5
#54 of 274 scored.
OpenAIAug 7, 2025Closed
Capabilities Index
Range undefined–undefined
150.0
Context
Takes text, image, file
400K
Input
Per 1M tokens
$1.25
Output
Per 1M tokens
$10
Benchmarks
How it scores.
- GPQA Diamond81.6%
59th of 186
- SWE-bench Verified73.6%
19th of 32
- FrontierMath55.4%
42nd of 81
- ARC-AGI-29.9%
45th of 79
- Humanity's Last Exam21.6%
13th of 40
- SimpleQA Verified50.1%
25th of 77
- Mock AIME91.4%
47th of 176
- Terminal-Bench49.6%
14th of 35
Everything else Epoch has for it
- MATH level 598.1%
- Fiction.LiveBench97.2%
- Aider polyglot88.0%
- Lech Mazur Writing86.0%
- DTBench84.5%
- GeoBench81.0%
- METR Time Horizons69.6%
- ARC-AGI65.7%
- WeirdML60.7%
- DeepResearch Bench49.6%
- VPCT49.0%
- SimpleBench48.0%
- LMCA47.0%
- GDPval34.8%
- Chess Puzzles33.7%
- Balrog32.8%
- FrontierMath-Tier-4-v2-Private21.9%
- ProofBench18.0%
- Mystery Game Puzzles15.2%
- EBR-bench12.7%
- GSO-Bench6.9%
- Remote Labor Index1.7%
On the index
Around it.
Sources