The shockwave caused by the Chinese AI startup DeepSeek in Silicon Valley and Wall Street is still being felt. Another tech giant, Alibaba, has launched a new version of its AI model on Wednesday, Tongyi Qianwen Qwen2.5-Max, which is said to surpass DeepSeek-V3.
Alibaba Cloud stated in an official announcement on its WeChat account: “Qwen 2.5-Max outperforms OpenAI and the most advanced open-source AI models from Meta, and is also better than the newly released DeepSeek-V3.”
Reports indicate that Alibaba’s new AI tools can be accessed via an API, or users can log into Qwen Chat to experience them, such as chatting directly with the model or using features like artifacts, search, etc.
Reuters reported and analyzed that DeepSeek’s rapid rise in the past three weeks has not only pressured foreign competitors but also domestic ones. The timing of Alibaba’s flagship Qianwen launch was particularly strategic, coinciding with the Lunar New Year (the first day of the lunar new year), when most people in China are returning home to reunite with their families for the holiday.
DeepSeek launched its AI assistant based on the DeepSeek-V3 model on January 10 and the R1 model on January 20, causing a stir in Silicon Valley and causing tech stocks to plummet. The low cost of use and development of this Chinese startup has led investors to question the big spending plans of leading AI companies in the U.S.
However, DeepSeek’s success has prompted domestic competitors to rush to upgrade their AI models.
Two days after the release of DeepSeek-R1, TikTok’s owner ByteDance released an update to its top AI model. The company claimed that the model outperformed OpenAI’s o1 (backed by Microsoft) in the AIME test, a standard used to evaluate the AI model’s ability to understand and respond to complex commands.
This mirrors DeepSeek’s claim that their R1 model can compare with OpenAI’s o1 on several performance benchmarks.
DeepSeek and Domestic Competitors
According to Reuters, DeepSeek-V2 – the predecessor to DeepSeek-V3 – sparked a domestic AI price war after its release in May of the previous year.
Since DeepSeek-V2 is open-source and extremely cheap, costing just 1 yuan (0.14 USD) per million tokens (the data unit that AI models process), Alibaba’s cloud division announced a price reduction of up to 97% for many of its model versions.
Other Chinese tech companies followed suit, including Baidu and Tencent, with Baidu releasing China’s first ChatGPT-like version in March 2023.
According to Reuters, DeepSeek’s mysterious founder, Liang Wenfeng, stated in a rare interview with Chinese media outlet Waves in July that the startup “is not concerned” with the price war and that achieving artificial general intelligence (AGI) is the company’s primary goal.
