← Retour au blog
tech 23 June 2026

Unsloth GLM-5.2 - How to Run Locally

Learn how to run Z.ai's GLM-5.2 model locally using Unsloth Studio, the leader in SOTA performance for coding and reasoning tasks.

Article inspired by the original source
Unsloth GLM-5.2 – How to Run Locally ↗ unsloth.ai

Introduction

Large language models (LLMs) have revolutionized the field of artificial intelligence. The release of Z.ai's GLM-5.2, an open model delivering state-of-the-art (SOTA) performance in coding, reasoning, and agentic tasks, marks a significant advancement. With 744 billion parameters, including 40 billion active ones, and a million-context window, GLM-5.2 is now accessible locally thanks to Unsloth's dynamic GGUFs. This article guides you through the process of running GLM-5.2 on your own hardware.

Why Run GLM-5.2 Locally?

Running an LLM like GLM-5.2 locally offers several advantages. It allows for complete customization, enhances data privacy, and in some cases, can reduce costs compared to using cloud services. Additionally, having direct access to a model of this scale enables faster experimentation and more frequent iterations, which are crucial for AI developers and researchers.

Hardware Requirements

Before diving into installation, it’s essential to ensure your hardware is ready. GLM-5.2 uses dynamic quantization to reduce memory footprint. The 2-bit model, for instance, occupies about 239 GB of disk space, ideal for a Mac with 256 GB of unified memory or a 24 GB GPU with 256 GB of RAM. Here is a table of memory requirements based on the level of quantization:

  • 1-bit: 223 GB
  • 2-bit: 245 GB
  • 3-bit: 290-360 GB
  • 4-bit: 372-475 GB
  • 5-bit: 570 GB
  • 8-bit: 810 GB

Ensure your total available memory, including VRAM and system RAM, comfortably exceeds the quantized model file size.

Installing GLM-5.2

Step 1: Prepare Your Environment

Start by installing the necessary dependencies. GLM-5.2 is compatible with Linux, macOS, and Windows platforms. Ensure that Python 3.8 or higher is installed, along with necessary libraries like PyTorch and TensorFlow for optimization.

Step 2: Download and Configure the Model

Download the GLM-5.2 model from Unsloth’s official site. Once downloaded, extract the files into a directory of your choice. Use the dynamic GGUFs to optimize the model based on your specific hardware requirements.

Step 3: Running in Unsloth Studio

Unsloth Studio provides an intuitive user interface to interact with GLM-5.2. Configure the model via the interface by adjusting the 'thinking mode' settings according to your task needs:

  • Non-Thinking: For simple, quick tasks.
  • High Thinking: For moderately complex tasks.
  • Max Thinking: For highly complex tasks.

Performance and Use Cases

GLM-5.2 stands out for its ability to compete with established models like Claude 4.8 Opus and GPT-5.5. Its applications range from artificial analysis to code generation, making it a versatile tool for developers.

Examples

  • Code Generation: Running GLM-5.2 to generate automated script solutions.
  • Data Analysis: Use for interpreting and predicting data trends.

Conclusion

GLM-5.2 represents a major breakthrough for those looking to incorporate advanced AI capabilities into their projects while maintaining local control. With proper installation and optimized configuration, it becomes an invaluable asset for any entrepreneur or developer.

Let's discuss your project in 15 minutes.

GLM-5.2 Unsloth AI models local deployment quantization
Deepthix newsletter · 100% AI · every Monday 8am

An AI agent reads tech for you.

Our AI agent scans ~200 sources per week and ships the best articles to your inbox Monday 8am. Free. One click to unsubscribe.

Visit the newsletter page →

Want to automate your operations?

Let's talk about your project in 15 minutes.

Book a call