On Wednesday, China’s technology company Alibaba unveiled a new version of its artificial intelligence model Qwen 2.5, which claims to outperform the highly acclaimed DeepSeek-V3.

The surprise release of Qwen 2.5-Max on the first day of the Lunar New Year, when most Chinese people are on holiday, suggests that DeepSeek's rapid ascent over the past three weeks has put significant pressure on its competitors both at home and abroad.

Alibaba’s cloud division has announced that its latest AI model, Qwen 2.5-Max, has demonstrated superior performance compared to several leading AI models, including GPT-4o, DeepSeek-V3 and Llama-3.1-405B.

In a statement posted on its official WeChat account, the company stated, "Qwen 2.5-Max outperforms … almost across the board GPT-4o, DeepSeek-V3 and Llama-3.1-405B," highlighting the model’s advancements over OpenAI and Meta’s most sophisticated open-source AI systems.

The introduction of DeepSeek’s AI assistant on January 10, powered by the DeepSeek-V3 model, along with the release of its R1 model on January 20, has made waves in Silicon Valley. The Chinese startup’s low development and usage costs have raised concerns among investors, leading to declining tech stock prices and sparking debates over the massive spending strategies adopted by major U.S. AI companies.

At the same time, DeepSeek’s rapid success has intensified competition within China’s AI industry, prompting rival firms to accelerate the development and enhancement of their own AI models to keep pace with the evolving market.

Two days after DeepSeek-R1 was launched, ByteDance, the parent company of TikTok, introduced an updated version of its flagship AI model. The company asserted that its model had surpassed OpenAI’s o1, which is backed by Microsoft, in AIME, a benchmark test designed to assess how effectively AI models interpret and respond to complex instructions.

ByteDance’s claim mirrored an earlier statement from DeepSeek, which had also suggested that its R1 model was on par with OpenAI’s o1 in multiple performance benchmarks.