Text model

GLM 5.3

GLM-5.3 is a large language model from Z.ai that reuses the GLM-5.2 base model, with gains coming entirely from added reinforcement learning post-training on coding and long-horizon agentic tasks. It supports three reasoning effort levels (low, high, max) and reaches open-weight state of the art on coding benchmarks such as Terminal-Bench 3.0 and Agents' Last Exam, plus strong vulnerability discovery and exploitation performance on cybersecurity benchmarks like CyberGym and ExploitBench.

Type
Text
Weights
Open
Downloads
1.4M
Likes
1.9k
Pipeline
text generation
License
other

Model hub

zai-org/GLM-5.3

Repo
zai-org/GLM-5.3
Library
transformers
Access
Public
Technical tags and languages
en zh transformers safetensors glm_moe_dsa text-generation conversational eval-results endpoints_compatible fp8
Open on Hugging Face

Papers

1

Related tools

2

More from Z.ai

13

About Z.ai

1

Z.ai develops the GLM family of foundation models and related generative-AI applications and developer services.

Tools using GLM 5.3

27