Chinese technology company Alibaba on Wednesday, January 29, released a new version of its artificial intelligence model Qwen 2.5.
The mentioned firm claims that the new version of its specified model is superior to DeepSeek-V3, which has become a sensation in the global AI industry and is being actively discussed in the current week.
A new version of Alibaba’s existing artificial intelligence model has been named Qwen 2.5-Max. It is worth noting that in this case, the date of the debut is very remarkable. A new version of the artificial intelligence model was released on the first day of the Lunar New Year. This is the first day of the holiday season in China when most locals are not working and are spending time with their families. The date of the debut indicates that the success of DeepSeek has become a significant factor in impacting other players in the Asian country’s artificial intelligence industry. It is highly likely that shortly, the Chinese area of machine intelligence will experience a process of increased competition. From the point of view of the logic of standard market tendencies, this is the natural reaction of players to the high results of one of them. It is also worth noting that DeepSeek-V3 has become a signal for the global artificial intelligence industry as a whole. In this case, in the practical plane, it challenged the dominance of US companies in the AI area. DeepSeek also showed by its example that huge investments are not necessary to develop highly efficient artificial intelligence systems.
Alibaba’s cloud unit said in an announcement posted on its official account on the WeChat digital platform, Qwen 2.5-Max outperforms virtual products of the corresponding category such as OpenAI’s ChatGPT-4o, DeepSeek-V3 and Meta’s Llama-3.1-405B by almost all indicators. In this case, it means the most advanced open-source models of artificial intelligence.
DeepSeek released its AI assistant on January 10th. This virtual product is powered by the DeepSeek-V3 model. Then, on January 20, the R1 model made its debut. This virtual product has become a sensation in Silicon Valley. Against the background of the success of DeepSeek, a sharp drop in the value of shares of technology companies was recorded. The low spending of the Chinese brand associated with the development of artificial intelligence led investors to question the expediency of huge financial injections into such projects. At the same time, the AI industry in the United States is operating under a strategy involving massive amounts of investment. Moreover, US machine intelligence developers have ambitious plans, the implementation of which is associated with large financial injections.
For the players of the Chinese artificial intelligence industry, the success of DeepSeek has become a kind of motivational factor. Local AI developers have become more active in upgrading their digital cognitive models. For example, two days after the debut of DeepSeek-R1, ByteDance, which ownership structure includes TikTok, released an update of its flagship artificial intelligence model. The company claims that its mentioned product outperforms OpenAI’s o1. In this case, the comparison criterion is AIME, a benchmark test that measures how well artificial intelligence models understand and respond to complex instructions. It is worth noting that DeepSeek has stated that its R1 model competes with OpenAI’s o1 on several performance benchmarks.
DeepSeek-V2, which is the predecessor of DeepSeek-V3, has provoked a kind of price war in the Chinese artificial intelligence market. This virtual product debuted in May last year. The mentioned open-source model of artificial intelligence was very cheap. In this case, 1 million tokens have cost 1 yuan ($0.14). It is worth clarifying that a token is a unit of data processed by an artificial intelligence model. Amid the debut of DeepSeek-V2, Alibaba’s cloud unit announced price cuts for several of its AI models up to 97%. It’s worth noting that other Chinese technology companies have followed suit. In this case, for example, it means Baidu, which presented the first local analog of OpenAI’s ChatGPT in March 2023, and Tencent, the most valuable Internet company in an Asian country.
Liang Wenfeng, DeepSeek’s enigmatic founder, said during a conversation with Chinese journalists last summer that the startup did not care about price wars. Also in this context, it was noted that the brand was focused on achieving artificial general intelligence (AGI).
OpenAI defines AGI as autonomous systems that are superior to humans in solving the most economically valuable tasks.
Currently, large Chinese technology companies such as Alibaba have hundreds of thousands of employees. DeepSeek operates as a research laboratory. The startup’s employees are mostly young graduates and doctoral students from leading universities in China.
Liang Wenfeng, while talking to media representatives, said that the largest technology companies in the Asian country may not be well suited for the future of the artificial intelligence industry. In this case, DeepSeek’s lean operation and loose management style were opposed to the high costs and top-down structures of the biggest firms.
Liang Wenfeng noted that large foundational models require continued innovation and technology giants’ capabilities have their limits.
