GLM-5

GLM-5: From Vibe Coding to Agentic Engineering

What is GLM-5?

GLM-5: From Vibe Coding to Agentic Engineering

Overview of GLM-5

What it is

GLM-5 is a series of open-weight large language models developed by Z.AI, designed for complex systems engineering and long-horizon agentic tasks. The series includes GLM-5, GLM-5.1, GLM-5.2, GLM-5.3, and GLM-5.3-Flash. GLM-5.3 uses the same base model as GLM-5.2, with all gains coming from post-training. GLM-5.3-Flash introduces a hybrid architecture combining sparse and linear attention, along with Manifold-Constrained Hyper-Connections (mHC), to improve efficiency. The models range from 320B to 744B parameters, with 18B to 40B active parameters.

Key features

  • Stronger coding: GLM-5.3 achieves a 50% improvement over GLM-5.2 on the in-house Z.ai Code Bench, and open-source SOTA on Terminal Bench 3.0 and Agents' Last Exam.
  • Emergent cyber capability: GLM-5.3 is state of the art on CyberGym for vulnerability discovery, and more than doubles GLM-5.2 on exploitation benchmarks.
  • Solid 1M-token context: GLM-5.2 delivers a stable 1M-token context for long-horizon work.
  • IndexShare architecture: GLM-5.2 reuses the same indexer across every four sparse attention layers, reducing per-token FLOPs by 2.9× at 1M context length.
  • Improved MTP layer: GLM-5.2's speculative decoding increases acceptance length by up to 20%.
  • Asynchronous RL infrastructure (slime): GLM-5 uses a novel asynchronous RL infrastructure to improve training throughput and efficiency.

Use cases

GLM-5 is purpose-built for complex systems engineering and long-horizon agentic tasks. It is designed for tasks such as frontend and backend development, long-term operational capability (e.g., running a simulated vending machine business over a one-year horizon), and real-world terminal tasks. The models are also suitable for reasoning, coding, and agentic tasks that require sustained optimization over hundreds of rounds and thousands of tool calls.

Getting started

See the repository README for instructions on serving GLM-5 series models locally using frameworks such as SGLang, vLLM, Transformers, KTransformers, and Unsloth. For deployment on the Ascend NPU platform, inference frameworks such as vLLM-Ascend, xLLM, and SGLang are supported.

License and pricing

The license is Apache-2.0. The model is open source.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/zai-orgglm-5.svg)](https://yunodb.com/tools/zai-orgglm-5?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/zai-orgglm-5?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/zai-orgglm-5.svg" alt="Featured on YunoDB" height="20"></a>