GLM-5.3-Flash vs GLM-5.3

Best cell per row highlighted. Null means undisclosed — never counted as zero.

Z.ai
GLM-5.3-FlashGA
8.6
AI Panel
Z.ai
GLM-5.3GA
8.7
AI Panel

Identity & lifecycle

ProviderZ.aiZ.ai
Family / tierGLM-5 · FlashGLM-5 · Flagship
StatusGAGA
Released2026-08-262026-08-14
Knowledge cutoff——

Architecture & context

Context window1M1M
Max output tokens131K131K
Input modalitiestext, image, videotext
Reasoning modealwaysalways
Open weightsYesYes

Pricing (per Mtok)

Input$0.15$1.4
Output$0.5$4.4
Cached input$0.03$0.26
Batch input——
Free tierNoNo

Speed

Output speed (tok/s)——
Time to first token (s)——
Latency tiermediummedium

Trust & deployment

Trains on inputsNoNo
LicenseMITcustom-GLM-5.3-License
Clouds—Mistral AI, Alibaba Cloud Model Studio
Compliance——

AI Panel scoring

Unified score8.68.7
Decision Maker8.58.5
Domain Strategist99
Finance Lead9.59.5
Domain Practitioner99
Power User8.58.5
Skeptic77.5
Value score9.89.6

Benchmarks

Terminal-Bench84.388.2
LMArena Elo14751483
LMArena Coding Elo15251524

Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.