log inget an API key
Platform benchmark

glm-5.2

zai · GPQA-Diamond · seed 42 · 2026-11-11 09:58 UTC

ResultsGPQA-Diamond · n=96 · seed 42
84.4%

84.4% correct · 89.6% answered on GPQA-Diamond

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

Score per $
0.58
Run cost
$1.46
Latency p50
16.5 s
Latency p95
406.1 s
Items
96/96 completed
Dropped
0

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