BEIJING: Chinese technology major Alibaba has unveiled an upgraded version of its AI model, Qwen 2.5, which it claims is more powerful than the widely praised DeepSeek-V3.

The release of Qwen 2.5-Max on the first day of the Lunar New Year is unusual, as most people in China are off work celebrating with their families. This timing suggests Alibaba is feeling the pressure from DeepSeek’s rapid rise in the AI industry over the past few weeks.

According to Alibaba’s cloud unit, Qwen 2.5-Max outperforms top AI models like OpenAI, DeepSeek-V3, and Llama-3.1-405B (Meta) in various benchmarks.

DeepSeek’s Impact on the AI Market

DeepSeek’s recent advancements have sent shockwaves through Silicon Valley, even leading to a drop in tech stock prices. The launch of DeepSeek’s V3 model on January 10, followed by the R1 model on January 20, has made investors rethink the high development costs of AI models in the U.S.

In China, DeepSeek’s success has sparked fierce competition. Just two days after DeepSeek released R1, ByteDance (owner of TikTok) introduced an updated AI model that it claims performs better than OpenAI’s o1 in understanding and responding to complex instructions.

The AI price war in China began last May when DeepSeek released V2, an open-source model priced at just 1 yuan (USD 0.14) per 1 million tokens. This forced Alibaba’s cloud unit to slash prices by up to 97 percent, with other companies like Baidu and Tencent following suit.

DeepSeek’s Vision: Beyond the Price Wars

Despite this aggressive competition, DeepSeek’s founder Liang Wenfeng insists that his company isn’t focused on price battles. In a rare interview last July, he stated that DeepSeek’s main goal is to achieve AGI (Artificial General Intelligence) i.e, a level of AI that surpasses humans in most valuable tasks.

Unlike large corporations like Alibaba, which have massive workforces, DeepSeek operates more like a research lab. It mainly employs young graduates and PhD students from top Chinese universities. Liang believes that big tech firms with high costs and rigid structures may struggle to keep up with the fast-changing AI world.