Chinese Firm Releases Major Open-Source AI Model
China's ZipuZZ.AI has released GLM-5, a 744-billion-parameter open-source AI model. The model reportedly outperforms Google’s Gemini 3 Pro and Grok 4 on several benchmarks and runs on domestic Huawei chips. The release could challenge U.S. dominance in foundational AI models by democratizing access to state-of-the-art technology.
- Zhipu AI was spun out of the prestigious Tsinghua University in 2019 and raised approximately $558 million in an initial public offering on the Hong Kong Stock Exchange in January 2026. - The model uses a Mixture-of-Experts (MoE) architecture, meaning only a fraction of its 744 billion parameters—around 44 billion—are active during inference, which significantly lowers the computational cost to run it. - GLM-5 is set to be released under the permissive MIT License, which allows for unrestricted commercial use, modification, and distribution of the model's weights. - Zhipu AI is part of a competitive field of Chinese AI firms, including Alibaba, Baichuan AI, Moonshot AI, and DeepSeek, that are increasingly releasing powerful open-source models. - On specific agentic and coding benchmarks, GLM-5 achieved a 77.8% score on SWE-bench Verified and ranked first among open-source models on Vending Bench 2, a test of long-term operational capability. - The company was added to the U.S. Commerce Department's export control
