Zhipu delivers the latest results of AI Agent

Author | Liu Baodan

Editor | Huang Yu

It has been two years since ChatGPT emerged, and AI large model technology has entered a new stage of development, with AI Agents becoming the new focus of industry competition. The domestic AI unicorn Zhipu has also taken the lead in disclosing its layout in Agent technology.

On November 29th, at the Zhipu Agent OpenDay, the company's CEO Zhang Peng announced the upgraded version of AutoGLM. This version can not only execute complex operational processes of over 50 steps and perform tasks across different applications but also supports custom phrases for long tasks and possesses proactive decision-making capabilities.

In addition, Zhipu AI launched a PC-based autonomous Agent—GLM-PC, which is positioned to become a comprehensive desktop assistant, currently covering functions such as meeting proxies, document processing, web searching, and summarization.

This upgrade marks Zhipu AI's further advancement in pushing large model technology towards more complex interactions with the physical world based on language interaction.

Zhipu was founded in 2019, stemming from the technology transfer of the Knowledge Engineering Laboratory of the Computer Science Department at Tsinghua University, and is the earliest AI startup among the "AI Six Little Tigers." In the two years leading up to the rise of ChatGPT, Zhipu AI had already fully invested in the research and development of large model technology and has now become one of the companies with the most comprehensive AI technology layout.

The rapid launch of AutoGLM is also attributed to Zhipu's forward-looking technological layout, including its technological accumulation in large language models, multimodal models, logical reasoning, and tool usage.

Zhang Peng stated that from the Agent Bench in April 2023 to the CogAgent model in August 2023, Zhipu has dedicated a year and a half to the research and development of AutoGLM and the CogAgent model for GLM-PC.

For Zhipu, AutoGLM and GLM-PC are important steps towards the company's AI intelligent operating system and are essential paths towards AGI.

Unlike GenAI, Agents are goal-driven, capable of fully executing workflows, adapting, learning, iterating, collaborating with other systems and humans, and completing tasks end-to-end. Zhang Peng believes that Agents can be seen as the prototype of the large model general operating system LLM-OS.

Zhang Peng stated, "At this stage, AutoGLM adds an execution scheduling layer between humans and applications, significantly changing the form of human-machine interaction. More importantly, we see the potential of LLM-OS, and based on the intelligent capabilities of large models, there is a future opportunity to achieve native human-machine interaction, bringing the paradigm of human-machine interaction to a new stage."

The industry generally believes that 2025 will be the year of the Agent explosion. Gartner recently listed agentic AI as one of the top ten technology trends for 2025 and predicts that by 2028, at least 15% of daily work decisions will be autonomously completed by agentic AI, while this figure is 0 in 2024 Industry leaders such as Jensen Huang, Robin Li, and Ultraman have publicly expressed optimism about its future development. Jensen Huang believes that AI Agents are the future trend of artificial intelligence development, with countries around the world competing to adopt AI to accelerate innovation and improve productivity. Soon, companies will have AI Agents working alongside teams.

According to data from marketsandmarkets, the global autonomous intelligent agent market was valued at USD 345 million in 2019 and is expected to reach USD 2.992 billion by 2024, with a CAGR of 54% during this period. The agency predicts that the global market size for autonomous artificial intelligence and autonomous intelligent agents will reach USD 28.5 billion by 2028, with a CAGR of 43%.

For Zhizhu, early layout of AI Agents has won valuable time advantages, which is crucial for an industry undergoing rapid technological iteration. However, AI Agents are still in the early stages, and the competitive pressure faced by Zhizhu AI should not be underestimated.

Zhang Peng admitted that the current capabilities of large models are still some distance away from truly replacing human office work. In the future, the GLM team will continue to accelerate the research and development of Agent model products.

The challenges are also evident, as global giants have begun to invest in AI Agents. In June, Apple showcased its latest AI achievement, Apple Intelligence, at its developer conference. Companies like Anthropic (Computer Use), Google (Jarvis), and OpenAI (Operator) have also made Agents a major focus for 2025.

Since the second half of the year, investor enthusiasm for the AI large model industry has significantly declined, and AI startups are generally facing commercialization pressures. In response, Zhang Peng stated that while commercialization is important, it is not the only goal; Zhizhu AI hopes to find a balance between technological investment and commercialization.

Zhang Peng emphasized, "We view issues more from a technology-oriented perspective, guided by the ultimate goal, rather than being limited to the rapid monetization of a single technology. Our ultimate aim is to help everyone effectively solve productivity problems, not just to achieve quick profits."

According to IT Juzi data, Zhizhu has completed a total of 10 rounds of financing since its establishment, with a current valuation of 26 billion yuan. This year alone, the company has experienced four rounds of financing, with investors including major companies like Alibaba and Tencent, as well as capital institutions like Sequoia and Hillhouse. The Beijing Artificial Intelligence Industry Fund, Tsinghua Holdings, and Zhongguancun Science City are also shareholders of Zhizhu.

AutoGLM is just the beginning. After more than five years of technological accumulation, Zhizhu is beginning to truly enter the stage of large-scale application. Although it is still far from the paradigm of operating computers and mobile phones with a single sentence, the exploration journey has already begun