Alibaba (9988.HK) launched Qwen 2.5-Max, which surpasses DeepSeek-V3 and OpenAI’s GPT-4o and Meta’s Llama-3.1-405B, according to the company. The announcement made on Lunar New Year’s first day reflects Chinese tech companies’ sense of urgency regarding DeepSeek’s fast growth.
DeepSeek’s AI models have disrupted the industry after the release of DeepSeek-V3 and R1, which created uncertainty for Silicon Valley technology leaders. The company’s low-cost development approach combined with efficient AI usage has led investors to question the high spending strategies of U.S. AI leaders. The domestic companies Alibaba and ByteDance reacted by speeding up their model development in response to DeepSeek’s success.
![](https://wonderfulengineering.com/wp-content/uploads/2025/01/image-378.png)
ByteDance released its flagship AI model after DeepSeek-R1 on January 20 and announced it surpassed OpenAI’s o1 in AIME benchmark tests. DeepSeek announced before that its R1 model delivered performance at a level equal to OpenAI’s O1 while China witnessed a growing AI competition.
Open-source development combined with cost-efficient strategies by DeepSeek has reshaped the market structure. The release of DeepSeek-V2 in May 2019 started a price competition when it offered AI model usage at 1 yuan ($0.14) per 1 million tokens. Alibaba Cloud faced pressure to reduce prices by 97% following DeepSeek’s announcement, which led other Chinese tech giants Baidu and Tencent to do the same.
DeepSeek functions as a research laboratory within Alibaba’s organization, although it runs independently from the company’s extensive employee base. It is led by young graduates and PhDs who graduated from China’s top universities. According to Liang Wenfeng, who founded DeepSeek, the large corporations face challenges in innovation because of their bureaucratic systems and expensive operations. The DeepSeek research team focuses on developing artificial general intelligence (AGI) technology, which represents a potential transformation of AI capabilities.