Alibaba unveils Qwen 2.5 AI model amid rising competition from DeepSeek

DeepSeek's rapid rise has unsettled its domestic rivals

Chinese tech giant Alibaba has unveiled an updated version of its AI model, Qwen 2.5, claiming it surpasses the widely recognised DeepSeek-V3.

The release comes amid mounting pressure on Alibaba from domestic AI startup DeepSeek, which has rapidly gained prominence in recent weeks.

Alibaba’s cloud division said in an announcement on its official WeChat account:

"Qwen 2.5-Max outperforms almost across the board GPT-4o, DeepSeek-V3 and Llama-3.1-405B.” The company’s reference to OpenAI and Meta’s latest open-source AI models highlights its ambition to compete at the highest level.

DeepSeek's rapid rise has unsettled both Silicon Valley and its domestic rivals. Its AI assistant, powered by the DeepSeek-V3 model and released on 10th January, followed by the launch of its R1 model on 20th January, has drawn attention globally.

The startup’s reportedly low development and operational costs have led investors to question the vast expenditures of leading AI firms in the United States.

DeepSeek’s success has also triggered a wave of activity among Chinese AI firms as they race to enhance their own models. Just two days after the launch of DeepSeek-R1, TikTok parent ByteDance introduced an update to its flagship AI model, claiming it outperformed OpenAI’s o1 in AIME, a benchmark test assessing AI comprehension and responsiveness to complex instructions.

DeepSeek had previously asserted that its R1 model could rival OpenAI’s o1 on several performance metrics.

Rising competition in China

DeepSeek’s emergence has shaken China’s AI landscape. Despite this turbulence, DeepSeek's founder, Liang Wenfeng, remains undeterred. In a rare interview with Chinese media outlet Waves in July, Liang said the company was not concerned with price wars, as its primary objective was to achieve artificial general intelligence (AGI).

Unlike Alibaba, which employs hundreds of thousands of staff, DeepSeek operates more like an AI research lab, with a lean workforce consisting largely of young graduates and doctoral researchers from China’s top universities.

The news isn’t all going DeepSeek’s way, though. The company was the victim of a “large-scale malicious attack” earlier this week, and just today a New York-based cybersecurity firm found a trove of sensitive data from DeepSeek, unsecured on the open internet.