Inference
EN

GLM-5.2 vs GLM-5.3

Change models
GLM-5.2Z.AI
GLM-5.3Z.AI
IndexAll categories, out of 100.
60.6 ±3.8
65.8 ±3.5 best
GapIndex points below the top.
−5.2
Top
CoverageIndex weight that has results.
85%
85%
SpeedOutput tokens per second.
111/s best
86/s
EloLMArena rating and rank.
1469
1476 best
ReleasedMonth of release.
2026-06
2026-08
Input / 1MUS dollars per 1M input tokens.
<$0.01 best
$0.07
Output / 1MUS dollars per 1M output tokens.
$6.80
$1.40 best
Batch output / 1MBatch price per 1M output tokens.
$0.989
$0.989
1K chats2K tokens in, 500 out, each.
$3.41
$0.84 best
1K documents20K tokens in, 1K out, each.
$6.94
$2.80 best
1K agent tasks200K tokens in, 20K out, each.
$137.40
$41.98 best
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
61.7
68.2 best
CodingCode writing and repair.
63.5
65.7 best
ReasoningLogic problems and puzzles.
58.6
64.5 best
MultimodalTasks with images and text.
—
—
KnowledgeFacts and expert knowledge.
55.7
59.3 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
55.1
66.8 best
MathMath problems.
59.6
63.1 best

Benchmarks

49 shared
Only GLM-5.2: 28Only GLM-5.3: 47
GLM-5.2
GLM-5.3
Artificial Analysis Humanity's Last ExamKnowledge
41.1%42 / 222
42.3% best39 / 222
Artificial Analysis Long Context ReasoningReasoning
78.3%63 / 217
79.7% best45 / 217
GPQA diamond · Epoch AIKnowledge
91.9% best29 / 217
90.9%35 / 217
Critical Physics TasksReasoning
20.9% best26 / 214
19.1%32 / 214
OTIS Mock AIME 2024-2025Math
86.4%73 / 200
91.1% best58 / 200
Artificial Analysis GPQA DiamondKnowledge
89.5%43 / 199
91.7% best24 / 199
Artificial Analysis Omniscience AccuracyKnowledge
24.3%108 / 164
32.3% best73 / 164
LegalBenchLegal
84.1%39 / 149
84.8% best29 / 149
TaxEval v2Finance
73.3% best50 / 143
72.4%62 / 143
LiveCodeBenchCoding
69.5%92 / 142
80.5% best67 / 142
GPQA Diamond · Vals AIKnowledge
85.6%45 / 135
88.1% best35 / 135
MMLU ProKnowledge
86.7%39 / 135
86.8% best38 / 135
GDPval-AA normalizedAgentic
43.6%42 / 133
57.5% best12 / 133
Vibe Code Bench v1.1Coding
64.0%56 / 123
78.1% best36 / 123
Artificial Analysis SciCodeCoding
51.2%47 / 114
59.0% best12 / 114
MedScribeHealthcare
83.5%46 / 109
88.8% best11 / 109
MedCodeHealthcare
40.8%60 / 106
42.9% best49 / 106
SimpleQA VerifiedKnowledge
34.2%60 / 91
41.0% best51 / 91
Chess PuzzlesReasoning
22.0%57 / 90
37.0% best39 / 90
FrontierMath-Tiers-1-3-v2-PrivateMath
59.2%43 / 90
68.8% best31 / 90