Chinese AI lab Zhipu releases GLM-5 under MIT license, claims parity with top Western models

2026-02-12 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Advanced, medium

Summary

Chinese AI company Zhipu AI has released GLM-5, an open-source model with 744 billion parameters under an MIT license. The company claims GLM-5 matches Claude Opus 4.5 and GPT-5.2 on coding and agent tasks, featuring a Mixture-of-Experts architecture with 40 billion active parameters at any time and trained on 28.5 trillion tokens. It utilizes Deepseek Sparse Attention (DSA) to reduce deployment costs and supports document generation into .docx, .pdf, and .xlsx files. Crucially, GLM-5 runs on both Nvidia GPUs and Chinese chips from Huawei, Moore Threads, and Cambricon, addressing US export restrictions. The model also integrates with frameworks like OpenClaw and popular coding agents, and its release significantly narrows the performance gap with Western counterparts.

Key takeaway

For NLP Engineers and CTOs evaluating large language models for deployment, GLM-5 offers a compelling open-source option with competitive performance on agent and coding tasks. Its compatibility with Chinese domestic hardware like Huawei Ascend chips provides a strategic advantage for operations in regions affected by US export controls. Consider testing GLM-5 for applications requiring document generation or complex agentic workflows, especially if hardware flexibility is a priority.

Key insights

GLM-5 is an open-source, 744B-parameter Chinese AI model claiming parity with top Western models on agent and coding tasks.

Principles

• Mixture-of-Experts architectures scale model size efficiently.
• Hardware diversification is critical for geopolitical resilience.

Method

GLM-5 uses a Mixture-of-Experts architecture, Deepseek Sparse Attention, and an asynchronous reinforcement learning framework (slime) for training, supporting vLLM and SGLang for inference.

In practice

• Deploy GLM-5 on Huawei Ascend or other Chinese chips.
• Use GLM-5 for direct document generation (e.g., .docx, .pdf).
• Integrate GLM-5 with OpenClaw for cross-app workflows.

Topics

• GLM-5
• Mixture-of-Experts
• AI Agents
• Open-source Models
• Chinese AI Hardware

Code references

• zai-org/GLM-5
• THUDM/slime

Best for: NLP Engineer, CTO, VP of Engineering/Data, AI Engineer, Machine Learning Engineer, AI Researcher

Related on AIssential

• Zhipu AI's GLM-5.2 closes in on closed-source leaders in coding marathons — Zhipu AI's GLM-5.2 sets a new open-source benchmark for long-horizon coding with a 1-million-token context and novel architectural efficiencies.
• 😸 Anthropic’s $30B War Chest, the $200M Political Battle, and the Flood of Free Chinese Models — The AI landscape is marked by massive investments, political influence, and rapid advancements in both open-source and proprietary models.
• Z.ai GLM-5: New SOTA Open Weights LLM — Chinese AI labs are rapidly advancing open-weight models, emphasizing cost-efficiency, long-context, and agentic capabilities.
• z.ai's open source GLM-5 achieves record low hallucination rate and leverages new RL 'slime' technique — GLM-5 sets new benchmarks for open-source LLMs in hallucination reduction and agentic document generation at a competitive price.
• Insane Open Source AI Model Just Dropped — GLM 5.1 is an open-source model matching proprietary LLMs in software engineering benchmarks.
• GLM-5.2 is the step change for open agents — GLM-5.2 is the first open-weight model to credibly rival frontier closed models as a general agent, narrowing the capabilities gap.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.