Back to glm-4-5
Zhipu AI /

GL
GLM 4.5

Open WeightsChat & ReasoningtextUpdated July 28, 2025
Some details or benchmark scores on this page are self-reported by developers and unconfirmed.

GLM 4.5: Zhipu AI's Dual-Mode Reasoning MoE Model

Model Overview

GLM 4.5 is Zhipu AI's flagship open-weights Mixture-of-Experts language model, released in July 2025. With 355 billion total parameters and 32 billion active during inference, GLM 4.5 features a dual-mode reasoning system supporting both fast non-thinking responses and deep agentic thinking modes. It unifies reasoning, coding, and autonomous action capabilities with native tool use and multi-agent workflow support, all under an MIT license.


✨ Key Features

Specification Table
FeatureDescription
Dual-Mode ReasoningSupports both fast non-thinking mode and deep agentic thinking mode
355B MoE (32B Active)Efficient MoE architecture — frontier capability at reduced inference cost
128K ContextExtended context window for complex multi-turn tasks
Native Tool UseBuilt-in tool use and function calling capabilities
Multi-Agent SupportNative support for multi-agent workflow orchestration

🔗 Resources

Specification Table
ResourceLink
Zhipu AIz.ai
Blogz.ai/blog/glm-4.5

📜 License

MIT License — Fully open-source for commercial and non-commercial use.

Key Features

Mixture-of-Experts (MoE) architecture with 355B total and 32B active parameters

Feature 01

Dual-mode reasoning supporting both fast non-thinking and deep agentic thinking modes

Feature 02

Native tool use and multi-agent workflow support

Feature 03

You might also want to compare

Verified Sources

Tags

moereasoningthinking

Model Specs

open-weights

Parameters

355B (32B active)

Context Window

128K tokens

License

MIT

Deployment

self-hostable

Resources & Links

Lineage

Model Family

Part of the glm-4-5 family

Only release in this line currently tracked.

Curator Notes

Zhipu AI (z.ai) flagship MoE model, launching the thinking/non-thinking reasoning system.

Compare Specs

Compare parameters, context windows, modalities, and benchmark scores of this model side-by-side with others.

Compare Model