Zhipu AImodel systemcard
GLM-5.2
#10of 27 · general
5.0/5.1 point releases collapsed into the 5.2 flagship
Standing
dot = this model · shaded = the field's middle 60% · track = full fieldGeneral
55.2#10/27
Math
76.7#4/27
Coding
65.1#6/27
Health
77.5#3/26
Engineering
43.8#10/17
Cyber
69.2#5/14
Legal
50.2#12/26
Instruct
43.7#13/27
Agents
58.3#11/27
Professional
56.5#9/27
Hallucination
37.3#12/27
Speed
66.9#4/22
Value
56.1#8/27
Open
60.1#2/11
Measured
≈ imputed · amber dot = lab-self-reported among the sourcesGPQA Diamond90.3%#17
HLE47.9%#7
MMLU-Pro86.9%#15
AIME 202591.9%#4
LiveBench71.7%#18
SWE-bench V78.7%#13
LiveCodeBench81.9%#19
SciCode50.5%#16
CritPt20.9%#7
Terminal-Bench69.0%#15
APEX Agents34.7%#9
GDPval50.3%#12
Arena1469#11
AA Index52.6#10
Vals Index65.0#11
Epoch Capabilities Index152.2#16
Operations
$1.4 / $4.4$ per 1M tokens, in / out
$2.15blended
136tokens / sec
$0.19$ per solved task
1469arena elo · 19,995 votes