The China Alibaba technology company on Wednesday launched a new version of its artificial intelligence model QWEN 2.5 that, he said, surpasses the acclaimed Deepseek-V3.
The unusual moment of the presentation of Qwen 2.5-Max, the first day of the Lunar New Year, when most Chinese do not work and are with their families, it points to the pressure that the meteoric rise of the Chinese artificial intelligence company Deepseek In the last three weeks he has exercised not only on his foreign rivals, but also on his national competition.
“Qwen 2.5-Max exceeds GPT-4O, Deepseek-V3 and call-3.1-405b,” says Alibaba’s cloud unit in an advertisement published in its official Wechat account, in reference to the code code models Open Open Openai and Meta.
The release on January 10 of the Deepseek’s assistant, based on the Deepseek-V3 model, as well as the release of January 20 of their R1 model, have shocked Silicon Valley and have caused the collapse of technological actions. The supposed supposed costs of development and use of the Chinese emerging company have led investors to question the huge expense plans of the main companies of the United States.
But Deepseek’s success has also caused a struggle among its national competitors to improve its own AI models.
Two days after the launch of Deepseek-R1, Bytedance, owner of Tiktok, published an update of his flagship model, which, he said, surpassed O1 of OpenAi, backed by Microsoft, in Aime, a reference test that measures measuring The ability of AI models to understand and respond to complex instructions.
Deepseek states that its R1 model rivals OPENAI O1 in several performance tests.
You may be interested: We introduce you to Liang Wenfeng, the founder of Deepseek, one of the architects of the AI revolution
Deepseek against national competitors
The predecessor of Deepseek’s V3 model, Deepseek-V2, triggered a War of AI model prices in China after its launch last May.
The fact that Deepseek-V2 out of open source and had an unprecedented price, only 1 yuan ($ 0.14) for 1 million tokens-or data units processed by the AI model-led to the cloud in the cloud Alibaba to announce price cuts of up to 97% in a series of models.
Other Chinese technology companies followed their example, such as Baidu, which launched the first Chinese equivalent to Chatgpt in March 2023, and the most valuable Internet company in the country, Tencent.
Liang Wenfeng, the enigmatic founder of Deepseek, said in July, in a rare interview with the Chinese media Waves, that the emerging company “did not care” the price war and that its main objective was to achieve the AGI (general artificial intelligence ).
OpenAI defines AGI as autonomous systems that exceed humans in most economic value tasks.
While large Chinese technology companies such as Alibaba have hundreds of thousands of employees, Depseek functions as a research laboratory, mainly formed by young graduates and doctoral students from the best Chinese universities.
Liang said in his July interview that he believed that the largest Chinese technology companies may not be adequate for the future of the AI industry, contrasting their high costs and hierarchical structures with the agile functioning and the flexible management style of Deepseek.
“The great foundational models require continuous innovation, the capacities of technological giants have their limits,” he said.
With Reuters information.
Do you use more Facebook? Let us like to be informed
