Inference
EN

GLM-5.2 vs GPT-5.3 Codex

Change models
GLM-5.2Z.AI
GPT-5.3 CodexOpenAI
IndexAll categories, out of 100.
60.7 ±3.7
61.6 ±5.2 best
GapIndex points below the top.
−0.9 · no clear gap
Top
CoverageIndex weight that has results.
85%
85%
SpeedOutput tokens per second.
111/s best
52/s
EloLMArena rating and rank.
1469
—
ReleasedMonth of release.
2026-06
2026-02
Input / 1MUS dollars per 1M input tokens.
<$0.01 best
$1.75
Output / 1MUS dollars per 1M output tokens.
$6.80 best
$14.00
Batch output / 1MBatch price per 1M output tokens.
$0.989
—
1K chats2K tokens in, 500 out, each.
$3.41 best
$10.50
1K documents20K tokens in, 1K out, each.
$6.94 best
$49.00
1K agent tasks200K tokens in, 20K out, each.
$137.40 best
$630.00
ContextMaximum tokens in one request.
1.05M best
400K
AgenticMulti-step tasks with tools.
61.6 best
60.5
CodingCode writing and repair.
64.2 best
60.8
ReasoningLogic problems and puzzles.
58.4
66.9 best
MultimodalTasks with images and text.
—
57.2
KnowledgeFacts and expert knowledge.
55.5
63.9 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
55.0
59.4 best
MathMath problems.
59.5
—

Benchmarks

12 shared
Only GLM-5.2: 67Only GPT-5.3 Codex: 10
GLM-5.2
GPT-5.3 Codex
Artificial Analysis Humanity's Last ExamKnowledge
41.1%42 / 222
42.5% best38 / 222
Artificial Analysis Long Context ReasoningReasoning
78.3%63 / 217
83.3% best13 / 217
Critical Physics TasksReasoning
20.9% best26 / 214
16.9%43 / 214
Artificial Analysis GPQA DiamondKnowledge
89.5%43 / 199
91.5% best25 / 199
τ²-Bench Tool-Agent-User EvaluationAgentic
99.1% best1 / 149
86.0%54 / 149
Artificial Analysis IFBenchInstruction
73.3%29 / 146
75.4% best23 / 146
LiveCodeBenchCoding
69.5%92 / 142
87.3% best16 / 142
Vibe Code Bench v1.1Coding
64.0% best56 / 123
61.8%58 / 123
SWE-benchCoding
82.8% best18 / 87
78.0%33 / 87
SWE-bench ProCoding
62.1% best26 / 83
56.8%46 / 83
SWE-Bench verified · Epoch AICoding
78.7% best4 / 32
74.8%15 / 32
PostTrainBench v1 (archived)Coding
34.3% best1 / 13
17.8%11 / 13