China’s emboldened AI industry releases flurry of model updates

Stay in view of the free updates

Simply subscribe to artificial intelligence Myft Digest – it is delivered directly to your inbox.

Chinese artificial intelligence groups were rushing into the updates of the model before the Lunar New Year, as the world wakes up to the main progress of the sector led by Deepseek in the face of American chips restrictions.

On Monday, on the eve of the most important annual holiday in China, Hangzhou has released a new open -source model, which enhances its reputation as the head of the formerly dominated by American giants. The atmosphere has become hot in the aftermath of the typical releases of the ALIBABA technology giant, the moonshot and ZHIPU.

This is the equivalent of a massive release on Christmas Eve. “We have all worked additional work to get things before the holiday,” said Grand Grand Product Director.

While Deepseek’s completion of panic in the United States about the progress made by Chinese laboratories on alignment budgets, those familiar with the industry say it feeds on new “confidence” in China will stimulate investment.

Deepseek made faster progress than other Chinese model companies. “This gives them confidence that they are able to catch up with,” said China’s artificial intelligence investor.

Deepseek has acquired the world’s attention through a series of typical versions that show similar performance to those of our competitors such as Openai and Meta, although it claims to have a small part of computing resources and was banned from obtaining the latest NVIDIA processors by export restrictions. Last week, it released its R1 thinking model, an advanced model that Openai’s O1 can automatically learn and improve itself without human supervision.

“Depsik has achieved a lot of energy in artificial intelligence players in China, and more broadly in a broader scale in the global artificial intelligence community, which will use its results from its R1 paper to achieve progress in thinking models “. The research center embraces the face.

This week, investors threw shares related to the prosecution, as NVIDIA lost about 600 billion dollars in the market value on Monday. They were interacting with Chinese breakthroughs, which turned out to be constructed with strong models with a different strategy of the United States by building worrying computing groups in order to move forward in the artificial intelligence race.

On Monday, QWEN of QWEN of QWEN2.5-1M released a series of new models capable of dealing with the longest inputs, an important development that means that the model can be published for artificial intelligence agent applications with higher memory requirements, according to Wang.

On the same day, she released Deepseek Janus-PRO, a model to generate text into a picture that could exceed a state of competitors such as Openai’s Dall-E3 and AI’s stability for stable spread 3 on some criteria.

Zhipu, whose value in the last financing round in December, issued a price of $ 3 billion, last week, a update of the GLM-PC. The AI’s agent model aims to institutional customers, enabling computers to automatically complete tasks such as filling models or digesting financial reports.

Although ZHIPU has not freed a lot of interest in the development of LLM, it has progress among local startups in marketing their technology, with the support of local governments and state -owned institutions that have partnered with the company based in Beijing to publish their models.

Last week, the other Beijing -based Moonshot company, which has the famous AI Chatbot Kimi, has updated its thinking model to the Kimi K1.5, indicating strong results compared to the modified artificial intelligence models of the complex thinking tasks. The latest version can process text and images while dealing with long and complex queries.

One of the usual practices of Chinese technology companies is to issue products before a long vacation, with additional benefit that potential customers who have a lot of spare time during the break will test and explore.

Once the Chinese artificial intelligence players return from a break, the race continues to become the main player who develops artificial intelligence applications for commercial use. “If artificial intelligence agents can create an exciting commercial value, one or two LLM players have an opportunity to convert into a new generation of software companies,” said artificial intelligence investor.