Platform benchmark

claude-haiku-4-5

anthropic · GPQA-Diamond · seed 42 · 2026-07-15 23:42 UTC

ResultsGPQA-Diamond · n=96 · seed 42
59.4%

59.4% correct · 100.0% answered on GPQA-Diamond

Score per $
1.55
Run cost
$0.38
Latency p50
8.8 s
Latency p95
13.2 s
Items
96/96 completed
Dropped
0