Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Chinese artificial intelligence start-ups are overhauling the work models to fight for competitiveness after the widespread dissemination of Rival DeepSeek technology in the country.
Zhipu once thought that he was the beginning of China’s most prominent large language model, for two people who are familiar with this issue, because he pays attention to the establishment of sales work, he closed his hopes for the initial public.
Between China Other leading generative AI Startups, 01.AI, DeepSEEK have stopped a large language models to sell special AI work solutions; Baichuan, the health market has chosen to concentrate in the health market; Moonshot reduced the marketing budget for Chatbot to pay attention to model training.
People close to these companies that do not reject or respond to the comments show how the turns change the shape of Deepseek’s growth of China AI The industry.
Since the opening of the R1 model in late January, Hangzhou-based starting, by Beijing quickly put and saw the country’s AI champion Lightning setting From hospitals to local governments everywhere.
This has left the country’s best AI beginnings – in the last two years, as part of the AI BOOM, as part of the AI Boom – to re-evaluate the existing strategies with an effort to repeat the success of their existing strategies.
“China’s LLM market is rapidly strengthened around a handful of leaders,” An engineer in the research center, an engineer, “DeepSEEK asked for more resources than the development of many companies.”
Beijing capitalist 01.ai, enterprise capitalist and Google China stopped working in the work called “DeepSeek Age” built by the former head of Kai-Fu Lee.
A number of open source models of Yi use mass information sets for training in late 2024, using mass information sets to develop models of open source models, developers, developers and more powerful models. With the agreement with Alibaba, its foundation model, the team was transferred to the Internet giant due to the people familiar with the issue.
Last week, 01.aay, Deepseek’s models will be able to sell the AI solution to companies that want to place the models. 01.AI, a method used by DeepSeek to prepare their experience, models of “mixture of experts” called “mixture of experts”.
An approach to a large-scale database that broke information from the Internet and other sources, connects many small models taught to industrial specific information. Moe structure allows you to grow larger models of poorer components of poor companies, but it can be more difficult for third party developers.
DeepSeek, decided Focus on the study Instead of selling applications to companies, he left a space to be filled by intermediaries as 01.AI. The Internet giant Baidu also pivinted to offer the same service in recent weeks.
Moonshot, last year, his viral AI attracted attention to as a chatBot, but his popularity is often out of the opponents and starting opponents for competition.
In recent weeks, it reduced the marketing costs for the beginning, because it says the two people who are familiar with the problem, and increase the model’s preparation for Deepseek’s uninterrupted success and improve His Chatbot’s activities.
However, as soon as other applications, Moonshot is subject to an indefinite future because it works with cash without a fixed income. Starting users wanted to invite the Chatbot to make money to send virtual gifts to the AI character behind.
Last year, more than $ 1.3 billion with a total of $ 1.3 billion, with a mixture of calculating loans from the Chinese technological giant Alibaba and enterprise capital companies, the report of those familiar with transactions.
In early 2024, Alibaba considered Moonshot’s potential purchase target and provided the beginning of the $ 800 million investment in any future sales. In recent months, Alibaba, initial investments, founder Jack MA, the President of the President Eddie Wu directed to direct the Internal AI efforts. The change is unlikely to try to get Alibaban as in the future.
Beijing Basic Beginner Baichuan, after watching the consumer, AI Chatbots and enterprise work for education, financial and healthcare companies.
In February, Baichuan fired a special financial AI application to banks and investment funds and finished the finished sales group, and the end line ended, two people who are familiar with this issue, he said.
In this case, the company’s management announced that the employees are aimed at developing technology for hospitals that are diagnosed.
On the contrary, ZHipu, built by a prominent computer scientist of Tsinghua University, is still a large number of working line. This has launched a number of consumer applications, as well as an enterprise for individual competitions and companies, formal and low margin businesses to local governments and companies.
The starting enterprise is burning with cash because it builds a sales job. In 2024, ZHipu, according to three investors, according to 7 investors, RMB2bn RMB300mn ($ 41 million) was RMB2BN.
Ballooning costs showed concern after demonstrating a way to build advanced models in a smaller budget after some investors. Unlike the small workforce of about 160 employees, ZHipu employs about 800 people, the beginning of the largest LLM.
Zhipu hopes that in two people who are familiar with the issue, Beyhin hopes to increase cash after receiving one of the letter of Coveted for an IPO. The company must be approved by the regulators before the Tech-Focused Star Innovation Council can take a list.
Deepseek received the beginning of Beijing NOD before changing the competitive view of EU players in China. Zhipu said that earlier aims to list investors until the end of the year, he said that the two people knowing the issue. However, they added that the developments of the DeepSek may affect it, progress with an IPO.
Investors in the company, the Government’s Deepseek’s well Quiphu’nun Zhipu’nun Zhipu’nun Zhipu’nu’nun escalated to local governments, he said concerns that he could threaten the AI solution.
However, Deepselaek rocked the AI race in China, and took some rivals to decide whether the group accepted open source models to focus on a small potential market.
“By accepting the highest level models, companies can eliminate the need to invest in ten million dollars in the ratio of millions of dollars.”
