Zhipu AI, a leading Chinese AI company specializing in large language models (LLMs), has recently completed a new funding round with a pre-investment valuation of 20 billion yuan (US$2.8 billion).
This round was led by Zhongguancun Science City Ltd., raising several billion yuan, according to Chinese media reports. It’s the third financing round for Zhipu AI since 2024, highlighting the company’s rapid growth and increasing valuation.
In June 2024, Saudi Aramco’s venture capital arm, Prosperity7, was reported to have invested US$400 million in Zhipu AI. This significant investment propelled Zhipu AI to 20 billion yuan club.
As of now, the investor list of Zhipu AI, which is often referred to as China’s answer to OpenAI, includes prominent firms like Hillhouse Capital, Qiming Venture Partners, Sequoia Capital, and major internet companies such as Meituan, Ant Group, Alibaba, and Tencent. Additionally, government-backed funds like the Social Security Fund, Zhongguancun Fund, and Beijing Artificial Intelligence Industry Fund have also invested in Zhipu AI.
Founded in June 2019, Zhipu AI originated from the Knowledge Engineering Laboratory (KEG) at Tsinghua University. It stands out as a fully domestically funded and independently developed LLM company in China. Z
hipu AI has been a pioneer in the field, developing the GLM (General Language Model) pre-training architecture since 2020 and launching a 10 billion parameter model, GLM-10B. The company continued to innovate with the 130 billion parameter model GLM-130B in 2022, followed by the GLM-4 series in 2023, which includes open-source dialogue models and video generation tools.
Zhipu AI has openly positioned itself as a competitor to OpenAI. While Zhipu AI has made significant strides, it still faces a gap compared to OpenAI’s capabilities. For instance, Zhipu AI’s latest model, GLM-4, was touted as being on par with GPT-4, but in practice, it can reach about 90% of GPT-4’s performance.
Zhang Peng acknowledges that Chinese LLM models started later and face limitations in high-performance computing and data quality, creating a roughly one-year gap with global leaders.
OpenAIβs GPT series is known for its generality, portability, and scalability, making it applicable in various scenarios. Zhipu AI, on the other hand, follows a "large model + small model" approach, focusing on adapting to different use case scenarios and tasks through pre-training and fine-tuning.
In terms of user base, the gap is apparent. OpenAI’s ChatGPT surpassed 1 million users within five days of its launch in November 2022, and by January 2023, it had over 100 million monthly active users.
In contrast, Zhipu AI’s daily active users for its chatbot, Zhipu Qingyan, range between 100,000 and 400,000 as of November 2023.
OpenAI continues to widen the gap with recent releases like the o1 series models, which have shown significant improvements over GPT-4, particularly in areas like physics, biology, and chemistry.
The LLM industry has been experiencing an intense price war since May 2024, involving several key players. Zhipu AI reduced the price of its GLM-3Turbo model by over 80% this year. Other major Chinese tech companies like ByteDance, Alibaba, Tencent, and Baidu have also slashed prices for their AI models.
OpenAI has been a primary driver of this price competition, halving the price of its GPT-4o model in comparison to GPT-4-Turbo. While the ongoing price war could potentially accelerate AI adoption, it poses a challenge for companies like Zhipu AI, whose profitability is already limited.
Meanwhile, OpenAI is less affected due to its substantial revenue, which could reach five billion U.S. dollars this year, according to analyst estimates.
Zhipu AI is focusing on product and investment layouts to achieve commercialization. In 2024, the company plans to launch an open-source fund for large models, providing resources to support the open-source community and encourage innovation.
Additionally, Zhipu AI is expanding its ecosystem through the "Z Plan," a 1 billion yuan AGI ecosystem fund aimed at supporting early-stage projects in the large model space.
To date, Zhipu AI has invested in 11 companies, covering various sectors including AI model layers, intelligent legal services, and generative AI applications. This strategic investment approach aims to build a comprehensive large model ecosystem.
However, the key challenge remains: developing products that not only capture user imagination but also translate into tangible productivity.
Related News: