Platform benchmark

claude-opus-4-8

anthropic · GPQA-Diamond · seed 42 · 2026-07-21 02:02 UTC

ResultsGPQA-Diamond · n=96 · seed 42
86.5%

86.5% correct · 99.0% answered on GPQA-Diamond

Score per $
0.51
Run cost
$1.68
Latency p50
8.2 s
Latency p95
19.8 s
Items
96/96 completed
Dropped
0