Chinese tech company Alibaba 9988.HK released a new version of its Quven 2.5 Artificial Intelligence Model on Wednesday, claiming that it had crossed highly acclaimed Dipsek-V3. Unusual time of Cuven 2.5-Max release, the first day of the lunar New Year when most Chinese people are away from work and with their families, in the last three weeks, Chinese AI startups indicate the pressure of Deepsek’s meteorite growth in the last three weeks. Not only foreign rivals, but also its domestic competition.

Alibaba’s Cloud Unit said in a announcement posted on its official WeChat account, “QWEN 2.5-Max outparfarfor almost board GPT-4o, Deepsek-V3 and LLAMA-3.1-405B,” Alibaba’s cloud unit by Alibaba. In a announcement posted, the model said, referring to the most advanced open-source AI of Openai and Meta.

The release of AI Assistant of Deepsak, run by Deepsek-V3 model, released on January 10, as well as its R1 model released on January 20, shook the silicon Valley and dipped the tech shares, with the Chinese Startup There were reportedly low development and use costs. Inspired to question investors in the United States by leading AI firms to question the plan to spend huge spending.

But the success of Deepsek has also created a scuffle to upgrade its AI model among its domestic rivals.

Two days after the release of Deepsek-R1, Tikok’s owner Bidens released an update for its leading AI model, claiming that it performs Microsoft-back Openi’s O1 in a benchmark test, Which measures how AI models understand well and respond to complex instructions.

It echoed the claim of Deepsek that its R1 model rivals OPENAI’s O1 at several performance benchmarks.

Deepsek vs domestic competitive

The erstwhile of the V3 model of the Deepsek, Dipsek-V2 triggered the AI ​​Model Price War in China after the release last May.

The fact is that the Deepsek -V2 Open -Source and unprecedented was cheaper, only 1 yuan ($ 0.14) per 1 million tokens – or units of data processed by AI model – the cloud unit of Alibaba was announced, which was 97 The price of up to percentage was cut. A series of models.

Other Chinese tech companies followed the suit including Baidu 9888.HK, which released the first of China’s first mat in March 2023, China’s first mat, and the country’s most valuable internet company TENCENT 0700.hk.

Deepsek’s esoteric founder Liang Wenfeng said in a rare interview with Chinese media outlet waves in July that the startup “did not care about price wars” and achieving AGI (Artificial General Intelligence) was its main goal.

Openai AGI defines as autonomous systems that cross humans in most economically valuable tasks.

While there are hundreds of thousands of employees in large Chinese technology companies like Alibaba, Deepsek acts like a research lab, mainly staffed by young graduates and doctoral students of top Chinese universities.

Liang said in his July interview that he believed that China’s largest technical companies could not be well suited for the future of the AI ​​industry, their high cost and unlike top-down structures, lean operations of Deepsek And with loose management style.

“Large fundamental models require constant innovation, their limitations in technical giants’ capabilities,” he said.