Z.ai Unveils Cheaper GLM-4.5 Model to Rival DeepSeek

In a bold challenge to rivals China-based AI start-up Z.ai formerly Zhipu, has unveiled its latest AI model. The GLM-4.5 with a bravado that it will cost-effectively perform better than other AI models. The company announced it on Monday at the World AI Conference in Shanghai. 

Zhang Peng, the CEO of Z.ai, said that GLM-4.5 is not only cheaper than the DeepSeek model developed but also is more efficient. The model is constructed based on so-called agentic AI, which allows subdividing a task into smaller chunks to finish them with greater precision.

Advantages of Open-Source and Reduced Operating Costs:

The main feature that is outstanding with GLM-4.5 is that it is open source. This implies that the developers can use the model at will. The model is also lightweight; it only runs on eight Nvidia H20 chips that is half the size of the model of DeepSeek. 

This was because the H20 chips were custom-made to suit China after the U.S. imposed restrictions on exports although the supply of these chips can be delayed. Zhang said that Z.ai has sufficient computing power at present and currently there is no necessity to have more of these chips.

Lower Prices of Tokens as Compared to DeepSeek and Others:

The new pricing of Z.ai also takes the headlines. The company will be charging only 11cents per million input tokens as against DeepSeek R1, which charges 14cents. In the case of output tokens, Z.ai will operate at 28 cents per million, which is much cheaper than DeepSeek at $2.19 dollars. 

The entities that AI models represent data with and process to generate new data are called tokens. Such prices are also lower than those of the Kimi K2 model provided by Alibaba-supported Moonshot, with 15 cents and 2.50 input and output tokens, respectively.

Increasing Competition in the Sphere of Chinese AI companies:

The introduction of the AI model GLM-4.5 contributes to the increasing surge of AI models by Chinese tech firms. This was the case in the past few weeks when Tencent released the HunyuanWorld-1.0 in order to develop 3D game scenes.  Alibaba launched Qwen3-Coder, which dealt with computer code generation. These models are all open source, and a trend is very evident in terms of developing AI in China and making it easily accessible.

Growth and International Surveillance by Z.ai:

Since its establishment in 2019, Z.ai has expanded rapidly. It has received more than $1.5 billion in investment from leading funds like Alibaba, Tencent, Qiming Venture Partners, and Prosperity7 Ventures, a subsidiary of the Saudi Company, Saudi Aramco. The local governments in Hangzhou and Chengdu also support the company. It is reported that Z.ai is in the planning phase of an IPO in Greater China.

In spite of this achievement, the company remains under the watchful eye of the world’s competitors. OpenAI cited Zhipu in June in a warning of Chinese AI development. Z.ai is also on such a list, the list of trade restrictions prepared by the American government, which restricts its access to American technologies and collaborations.

Global Context: DeepSeek and United States apprehensions:

In January, DeepSeek was in the headlines worldwide due to the creation of a potent AI model that was designed despite the U.S. chip limitations. The company has asserted to have trained its V3 model at less than 6 million dollars. 

Nonetheless, analysts have mentioned that this expenditure did not signify the entire hardware investment, and that it was above $500 million dollars. The capability of DeepSeek to produce high-performance at low costs caused panic in Washington and other tech giants in the West. 

The AI-Rapid Race in China Carries on

This trend is demonstrated by the introduction of GLM-4.5, an AI system designed to help the Chinese achieve their ambitious nine-point goal within one year. Z.ai is not only keeping up with giants such as DeepSeek or OpenAI but also pushing to compete on their level. As well as increased firms selling open-source models and affordable prices, China is now starting to become a key player in the global race to become the most competitive and leading in Artificial Intelligence. What the world will do with these swift changes will be seen in the coming months, and the question is whether new regulations or innovations will be the outcomes of the same.