Zhipu AI

AsianFin — Zhipu AI, one of China's "Six Little Tigers of Large Models" with a valuation of 20 billion yuan, has completed a strategic funding transaction exceeding 1 billion yuan (US$ 138 million).

Investors in the transaction include Hangzhou Urban Investment Industrial Fund and Shangcheng Capital, marking the first time Hangzhou's state-owned capital has invested in a major AI model company.

The funding aims to accelerate technological innovation and ecosystem development for Zhipu's GLM (General Language Model) foundation model.

Zhipu has established Zhejiang Zhipu Xinpian Technology Co., Ltd. in Hangzhou, leveraging cutting-edge AI and large model technologies to better serve the thriving real economy in Zhejiang province and the Yangtze River Delta region. The company is committed to driving the digital transformation of industries through AI technology.

This transaction also represents Hangzhou state-owned capital's first foray into the AI large model sector, following its earlier support for the "Six Little Dragons of Hangzhou," which are Game Science, DeepSeek, Unitree Robotics, Deep Robotics, BrainCo, and ManyCore. It is also seen as a crucial strategic funding round for Zhipu ahead of its potential IPO.

Founded in 2019 with the vision of enabling machines to think like humans, Zhipu has developed the GLM pre-training architecture, which boasts independent intellectual property rights in China. Its latest foundation model, GLM-4-Plus, positions the company among the world's leading AI technology providers.

Since launching its open-source billion-parameter conversational model, ChatGLM, in March 2023, Zhipu has iterated through four major versions of the GLM model, releasing over 20 AI model technologies and products.

On the enterprise front, Zhipu AI had partnered with more than 2,000 ecosystem collaborators by the end of 2023, with over 1,000 large-scale model applications and deep collaborations with more than 200 enterprises. These partnerships span industries such as media, consulting, consumer goods, finance, new energy, internet, and smart office solutions.

Zhipu has outlined a five-stage development path to Artificial General Intelligence (AGI): L1 (pre-trained large models), L2 (alignment and reasoning), L3 (self-learning), L4 (self-awareness), and L5 (conscious intelligence).

The stage is currently between L2 to L3, with scaling laws remaining fundamental. The key to model performance lies in ultra-large-scale high-quality data, advanced algorithms, and enhanced computing power.

Looking ahead, 2025 is set to be Zhipu's "year of open source," with plans to release and open-source a new series of large models, including foundation models, reasoning models, multimodal models, and agent-based systems. This move is expected to further solidify Zhipu's position as a leader in China’s AI innovation.

(Note: 1 yuan equals US$ 0.1378)