How 360 and ChatGLM Are Building China’s “Microsoft + OpenAI” Large‑Model Duo

360 and Zhipu AI Join Forces

On May 16, 360 Group and Zhipu AI announced a strategic partnership to co‑develop a trillion‑parameter large model called “360GLM”, which the parties claim meets the standards of a next‑generation cognitive general‑purpose model.

Why the Collaboration Mirrors Microsoft + OpenAI

360 founder Zhou Hongyi has repeatedly pointed out that Microsoft complements OpenAI’s engineering, scenario, product, and commercialization capabilities, creating a “Microsoft‑OpenAI” division of labor that drove the recent AI inflection point. He argues that China should adopt a similar model of large tech firms partnering with leading research institutions to form a “Microsoft + OpenAI” combo for large‑model breakthroughs.

Zhou says the current cooperation with Zhipu AI follows this joint‑research‑development paradigm.

About Zhipu AI

Zhipu AI, a company commercializing Tsinghua University research, has developed the bilingual (Chinese‑English) dialogue robot ChatGLM. The ChatGLM project is in internal testing and aims to promote large‑model scientific research and support academic exchange and industry collaborations. The demo site is https://chatglm.cn.

360 Brain (360智脑)

At the 2023 Digital Security and Development Summit hosted by 360, founder Zhou unveiled “360 Brain”, a China‑style ChatGPT product based on a large language model. The product will first be deployed in search scenarios to provide more intelligent and efficient search for enterprise users.

Demonstrations show that the 360‑version GPT already exhibits some reasoning ability, but it sometimes fails to answer specific questions correctly, requiring multiple attempts for accurate responses. Enterprise users can apply for internal testing on the 360 search homepage, and future integrations include browsers, digital assistants, and intelligent marketing.

The launch follows similar releases from Baidu’s Wenxin Yiyan and Alibaba’s Tongyi Qianwen, further enriching China’s AI market.

Strategic Layout of 360GLM and 360GPT

Zhipu AI is regarded as a top domestic AI company, described as having the “most OpenAI‑like temperament”. In a Stanford Large‑Model Center evaluation of 30 mainstream models, Zhipu’s 130‑billion‑parameter GLM‑130B was the only Asian model selected, with accuracy and other key metrics comparable to OpenAI, Google Brain, Microsoft, and NVIDIA. Over 70 countries and 1,000 institutions have applied to use it. ChatGLM, built on GLM‑130B, aligns with human intent through supervised fine‑tuning, supports Nvidia, Huawei Ascend, HaiGuang, and ShenWei chips, and its open‑source ChatGLM‑6B has been downloaded over 1.6 million times, topping Hugging Face’s global model trend chart for two weeks.

Zhipu AI CEO Zhang Peng says the company’s vision is to let machines think like humans and realize “Model as a Service” (MaaS). 360 brings a strong multimodal R&D team and long‑term AI expertise, plus search and browser scenario advantages, making it a powerful partner for data, reinforcement learning, engineering optimization, user scenarios, and commercialization.

Through the partnership, 360 adopts a “dual‑engine” model of independent and collaborative R&D. Earlier in March, 360 unveiled its own trillion‑parameter model “360GPT”, which has achieved impressive results in intelligent search and AI‑generated image tests. 360GLM and 360GPT each have distinct strengths and complement each other; they will be seamlessly integrated at the application layer to deliver a smooth user experience. Zhipu AI will continue to deepen large‑model applications based on this cooperation.