GLM 5.3 Flash specs, prices, where it runs and its benchmark standing

Name: GLM 5.3 Flash
Author: Z.ai

open weights · top 10 on 2 of 10 · released Aug 20, 2026

Provider: Z.ai
Release date: Aug 20, 2026 · Pricing lists Aug 26, 2026; Boards
Parameters: 321B total; Hugging Face
Modalities: Text, Image, Video to Text; OpenRouter
License: MIT; Hugging Face
Context window: 1.05M tokens; Pricing
Pricing: $0.01 in / $0.16 out per 1M tokens; All prices
Hosting: 41 priced API routes · OpenRouter · self-host: weights on Hugging Face; Pricing
Hardware: Estimate: about 328 GB of weights as published (F8 E4M3), about 321 GB at 8-bit, about 161 GB at 4-bit, plus memory for the context; Hugging Face config
Weights: Open weights
Benchmarks: 10 boards captured

Rank	Board	Category	Score	Of
2	OpenRouterUsage · as “z-ai/glm-5.3-flash-20260826”	Usage	13.2T tok	18
5	AA SpeedSpeed · as “GLM-5.3-Flash” Unverified	Speed	39 tok/s	43
25	AA IntelligenceReasoning · as “GLM-5.3-Flash”	Reasoning	27.8	50
28	Text ArenaChat · as “GLM-5.3-Flash”	Chat	1,476	84
28	ARC-AGI-2Reasoning · as “GLM-5.3-Flash (Max)”	Reasoning	65.8%	92
28	WebDev ArenaCoding · as “GLM-5.3-Flash”	Coding	1,607	82
29	Vision ArenaVision · as “GLM-5.3-Flash”	Vision	1,272	85
43	GPQA DiamondReasoning · as “glm-5.3-flash_max”	Reasoning	90.2%	100
47	Epoch ECIReasoning · as “GLM-5.3-Flash”	Reasoning	153.6	100
51	LiveBenchReasoning · as “glm-5.3-flash”	Reasoning	71.6%	64