Shulou(Shulou.com)12/24 Report--

Recently, according to the third quarter financial report of Yuncong Technology, Yuncong Technology realized business revenue of 182 million yuan, an increase of 181.27% over the same period last year, of which the business revenue of realizing man-machine cooperation operating system was 212 million yuan, an increase of 126.40% over the same period last year and exceeding the level for the whole year of 2022. At the same time, the market capitalization of Yuncong Technology rose continuously in the following trading day, which attracted the attention of the market. To investigate the reason, the value of the cloud from the leisurely model is being reflected.

The continuous upgrading of the leisurely model promotes the intelligent reform of the industry.

The Yuncong leisurely model, launched in May this year, already had extremely strong product power at that time, with better interactive performance, and was used in industries such as government affairs, transportation, energy, education, entertainment and so on. In the same month, Yun also officially put into operation from the Western Intelligence Center, marking the official commissioning of the calm big model production line. The scale of the whole center has reached 1200PetaFlops in the near future and will reach 5000PetaFlops in the future. When it is fully launched, it will be the largest high-performance computing center in the west.

Two months after the first release of the large model, Yuncong updated version 1.5 of the leisurely large model. This version contains a variety of model specifications of billion, ten billion and hundreds of billions, and the measured performance ranks first in the global comprehensive examination and evaluation of large models (C-Eval), with a total list of 4. The focus of version 1.5 is on the industry model with 13 billion (13B) parameters, which is distilled from hundreds of billions of models. Among the 52 test indexes, 44 exceed ChatGPT and 29 exceed GPT-4.

In addition, Yuncong Technology joined hands with Jiadu Technology and Chongqing Urban Transportation Development Investment Group to release the large model of Zhixing city transportation industry, and jointly launched the joint research and development of port model PortGPT with Tianjin Port Group, Zhejiang Seaport Group, Huawei and Tianjin Supercomputing Center.

Large industry models such as transportation and ports can not only make a comprehensive and in-depth analysis and prediction of the urban traffic system, but also monitor the use of roads, public transport, lanes and other traffic facilities in real time. accurately assess traffic flow, congestion and travel demand, and provide a scientific and accurate decision-making basis for urban planners, policy makers and citizens. It can also serve port production, logistics services, customer management, customer service, video analysis, office support and other six major application scenarios.

Signing a comprehensive cooperation with Huawei is worth imagining in the future market.

On November 6, Yuncong Technology Chief Executive responded to questions from the investor platform that the company had signed a comprehensive cooperation framework agreement with Huawei. Pure domestic Investment Co., Ltd. has created a listed AI company in conjunction with national science and technology enterprises, which is undoubtedly the "Wang Fang combination" in China's high-tech field. In fact, before this, the two sides already had practical cooperation in the field of large models. On the day of the release of the large models calmly, Yuncong Technology and Huawei Teng established a large model ecological alliance to start the large model ecological cooperation.

On July 6, 2023, the two sides first launched the "leisurely large model training and reasoning integrated solution" at the World artificial Intelligence Conference (WAIC), which laid a solid foundation for the further promotion of the domestic large model ecosystem, the most important of which is the leisurely large model training and pushing all-in-one machine.

The training and push all-in-one machine is equipped with Teng AI hardware and distributed computing platform, stand-alone FP16 computing power up to 2.5PFLOPS, training performance 8000tokens / s, combined with the traditional visual advantages of Yuncong, it can provide language, visual and multi-modal basic model reasoning and training capabilities, and based on calm large model algorithms and tools, which greatly reduces the difficulty and cost of user training and managing large models. The training and push all-in-one machine realizes the training and reasoning process of seamlessly switching large models in the same cluster, and has built-in large model reasoning engine, tuning tool and operator acceleration library, which can be used immediately out of the box. The integration of cloud's long-term accumulated industry data and algorithms from science and technology, large transportation, intelligent manufacturing and other fields, as well as the technical advantages of Teng AI chip and AI framework, can accelerate the landing application of large models in various industries.

The person in charge of Yuncong science and technology told reporters that the way to cooperate with Huawei is to jointly build products. Huawei's contract manufacturer will integrate the chip with Yuncong's large model, and then jointly promote it to customers. In this cooperation, Yuncong will not only sell products, but also provide other services related to application development. The whole market space is different from the previous sharing model, but through joint promotion, jointly create products, and promoted by Huawei sales teams.

From the perspective of the overall market, on the one hand, it pays attention to the increase in demand for domestic Xinchuang and the deployment of compliance and privatization, and the other is to pay attention to data elements and changes on the application side based on large models. Driven by data security and data elements, the large model of AI brings the rigid demand of localized deployment of large models of central state-owned enterprises and government.

Huawei plans to sell 3000 all-in-one phones in 2024. At present, only iFLYTEK, Yuncong Technology and Zhisu are producing and selling these all-in-one phones. IFLYTEK's sales scope is limited to Jiangsu. Yuncong technology can be sold nationwide.

Based on the price range of 10-3 million per unit price and 10% market share, this will lead to explosive growth in revenue and net profit of 10-3 billion. The market space will also be expanded with the development of policy and technology.

In the blowout of large model in 2023, cloud successfully found the second growth curve from science and technology. Through the innovation and application of large model technology, it has excavated a trillion-level large model industrialization gold mine, which is expected to achieve sustained and rapid business growth.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.