log inget an API key
Platform benchmark

glm-5.3

zai · GPQA-Diamond · seed 42 · 2026-11-11 10:43 UTC

ResultsGPQA-Diamond · n=96 · seed 42
82.3%

82.3% correct · 86.5% answered on GPQA-Diamond

A low answered rate means the run starved on timeouts — not that the model is weak.

Score per $
0.22
Run cost
$3.77
Latency p50
108.5 s
Latency p95
714.4 s
Items
96/96 completed
Dropped
0

Dataset: GPQA-Diamond · question and response text withheld under dataset access terms.