Overview
GLM-4.7-Flash is a lightweight, open-weight GLM-4.7-series model intended to balance performance and deployment efficiency. It uses a mixture-of-experts architecture with approximately 30 billion total parameters and 3 billion active parameters. Its weights are published under the MIT license. (huggingface.co)
Release
Published on January 19, 2026. Zhipu’s release announcement describes it as a free model optimized for frequent use. (docs.bigmodel.cn)
Capabilities and uses
- Text input and text output; the domestic API documentation specifies a 200K-token context window and up to 128K output tokens.
- Supports thinking modes, streaming, function calling, context caching and structured JSON output. The domestic documentation also lists MCP integration.
- Intended uses include agentic coding, code generation, task planning, writing, translation, reasoning, role-play and long-text work. (docs.bigmodel.cn)
Limitations
The documented model interface is text-to-text; image, audio or video input should not be assumed. Generated code and answers require review. API availability and usage limits depend on the platform and account. Running the open weights locally requires suitable computing resources. (docs.bigmodel.cn)
Official API pricing
- International Z.AI API, USD: The official pricing table lists input, cached input, cached-input storage and output as Free for GLM-4.7-Flash; its token-price table uses a per-1-million-token unit. This is distinct from the paid GLM-4.7-FlashX model. Official international pricing, checked November 17, 2026. (docs.z.ai)
- Mainland China BigModel: The official release announcement identifies the model as free. The current domestic pricing page could not be read without JavaScript, so its detailed CNY input and output rates could not be independently verified. Checked November 17, 2026. (docs.bigmodel.cn)