Z.ai·GLM

GLM-5.3-Flash

The first natively multimodal GLM-5 Flash model: 320B total, 18B active parameters with hybrid sparse and linear attention.

CurrentReasoningCodingAgentMultimodalOpen weights

MODEL NUMBER083

Specifications

Only published specifications are shown.

Total parameters

320B

Active parameters

18B

Architecture

Hybrid

Context

Input modalities

Text · Image · Video

Open weights

Yes

License

MIT

Status

Current

The first natively multimodal GLM-5 Flash model: 320B total, 18B active parameters with hybrid sparse and linear attention.

It uses a new base model and 1M context to deliver GLM-5 long-horizon agent capability at lower inference cost.

Compared with GLM-5.3 ↗

PreviousGLM-5.3— · — ctx

CurrentGLM-5.3-Flash320B · 1M ctx

New capabilities Native multimodal Computer use

Sparse AttentionLinear Attention

Native multimodalLong-horizon agentCodingComputer use

2026-08-30Open weights

GLM-5.3-Flash launched with MIT-licensed open weights.

Select a card to view the related model.