Z.ai releases GLM-5.3-Flash with open weights and native multimodal input

GLM-5.3-Flash is a new 320B-parameter mixture-of-experts model with 18B active parameters, multimodal input, and an MIT license. Here is what changed and what to verify.

Quick answer

Z.ai released GLM-5.3-Flash on August 26, 2026 as the first natively multimodal model in the GLM-5 series. It is a 320B-parameter mixture-of-experts model with 18B active parameters, a hybrid sparse-and-linear-attention architecture, a published one-million-token context configuration, and weights under the MIT license. Availability in Chat AI has not been verified.

Download Chat AI Opens the official App Store or Google Play for your device.

What Z.ai released

GLM-5.3-Flash is a new open-weight mixture-of-experts model for text, image, coding, and agent workflows. Z.ai describes it as the first natively multimodal member of the GLM-5 series. The official model card lists 320B total parameters, 18B active parameters, and an MIT license.

Sources: Z.ai, Z.ai on Hugging Face

The architecture targets long-context efficiency

Z.ai combines linear attention for local dependencies with sparse attention for retrieving relevant global context. The release also uses Manifold-Constrained Hyper-Connections and an IndexPool mechanism that compresses indexer keys. Z.ai says these changes reduce attention compute and KV-cache size relative to GLM-5.3, but the practical benefit will depend on context length, hardware, quantization, and serving software.

Sources: Z.ai, GLM-5 Team

Multimodal input expands the coding loop

The release can use visual context alongside text, which Z.ai positions for tasks such as inspecting rendered interfaces, documents, spreadsheets, presentations, dashboards, and browser or computer-use workflows. That is broader than simply generating code from a text prompt. Teams should still test image detail, small text, tool reliability, and error recovery on their own material.

Sources: Z.ai, Z.ai on Hugging Face

One-million-token context is a supported configuration, not a quality guarantee

Z.ai documents support for context lengths up to one million tokens and explains several serving optimizations intended to control memory and latency at that scale. A large context window does not guarantee that every detail will be recalled or used correctly. Evaluate retrieval accuracy, latency, memory, and total cost with inputs that resemble the real workload.

Sources: Z.ai, Z.ai on Hugging Face

Weights, hosted access, and first-party benchmark claims

The weights are available on Hugging Face under the MIT license, and Z.ai lists SGLang, vLLM, and TokenSpeed among the supported local-serving options. Z.ai also rolled the model out to its GLM Coding Plan and ZCode products. Cloudflare subsequently added the model to Workers AI as @cf/zai-org/glm-5.3-flash, available through Workers bindings, REST, an OpenAI-compatible endpoint, or AI Gateway; Cloudflare says it requires a Workers Paid plan or prepaid AI Gateway credits. The release includes strong publisher-run coding, agentic, and vision results, so reproduce the tasks that matter before choosing a deployment.

Sources: Z.ai, Z.ai on Hugging Face, Cloudflare

GLM-5.3-Flash is not yet verified in Chat AI

GLM-5.3-Flash does not appear in Chat AI's public model directory at publication time, so this article does not claim that it is available in the app. Check the current directory and in-app model selector before assuming access; availability can vary by plan, platform, region, provider access, and app version.

Sources: Chat AI

Frequently asked questions

What readers usually ask

What is GLM-5.3-Flash?

GLM-5.3-Flash is Z.ai's natively multimodal, open-weight mixture-of-experts model with 320B total parameters and 18B active parameters. It is aimed at coding, agent, visual, and professional knowledge-work tasks.

Is GLM-5.3-Flash open source?

Z.ai publishes the model weights on Hugging Face under the MIT license. Review the model card, repository files, and dependency licenses for the exact deployment you plan to build.

Does GLM-5.3-Flash support images?

Yes. Z.ai describes it as the first natively multimodal model in the GLM-5 series and documents visual workflows alongside text, coding, and agent use cases.

Does GLM-5.3-Flash have a one-million-token context window?

Z.ai documents support for context lengths up to one million tokens. Actual memory use, latency, retrieval quality, and cost depend on the deployment and workload.

Is GLM-5.3-Flash available in Chat AI?

Availability in Chat AI was not verified at publication time. Check the current Chat AI model directory or in-app model selector before assuming access.

Evidence

Sources

  1. GLM-5.3-Flash release announcementZ.ai · Primary source
  2. GLM-5.3-Flash model card and weightsZ.ai on Hugging Face · Primary source
  3. GLM-5 technical reportGLM-5 Team · Primary source
  4. GLM-5.3-Flash availability on Workers AICloudflare · Primary source
  5. Chat AI model directoryChat AI · Primary source