the number of large model bidding projects continued to grow in august, reaching 127, with iflytek and baidu leading the industry
2024-09-04
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
in august, the number and amount of large model winning projects maintained a strong growth momentum, and the number and amount of winning projects in a single month continued to rise, proving that the commercialization of large models in the b-end market continued to accelerate. according to the "china large model winning project monitoring report (august 2024)" recently compiled by the media, there were 127 large model-related winning projects counted through public channels, of which 84 winning projects disclosed a total amount of 390 million yuan, setting a new monthly high since 2024.
as big models gradually penetrate deeper into industry scenarios and business levels, the number of application projects accounts for about 70%. science and education, central state-owned enterprises, operators and other fields have become key scenarios for the implementation of big models in august.
in the b-end market, iflytek and baidu lead the industry in the number of projects they have won. iflytek’s monthly winning amount exceeded rmb 100 million, ranking first in the industry.
data source: monitoring of china's large model winning projects from january to august (intelligent hyperparameter compilation)
the central state-owned enterprise market has become the main force in the implementation of large models, and application projects account for70%
according to the data of big model winning bids in august, the proportion of application projects has remained stable at around 70%. the types of bidding related to big models mainly include computing power, data, big models and applications, with application projects occupying a high position; in terms of landing scenarios, currently, they are mainly intelligent customer service, ai programming, digital humans, big model platforms, etc.
8the winning bids for large model projects in august were mainly concentrated in the fields of education, science, central state-owned enterprises, and operators. in august, the central state-owned enterprise field was the focus. the number of large model-related procurement projects of central state-owned enterprises accounted for 20%, but the amount accounted for about 42%, becoming the biggest driving force for the winning bid amount to reach a new high in a single month.
on the one hand, central state-owned enterprises are accelerating their digital transformation, responding to the "cloud-based, data-driven and intelligent" action in the 14th five-year plan, and realizing the "emergence of intelligence" in different business scenarios through large models. on the other hand, the implementation of large computing projects such as "east data west computing" provides a continuous source of computing power support for large model pre-training.
category and disclosed amount of large model bidding projects in august
in august, iflytek and baidu maintained their industry-leading number of winning projects. iflytek's winning bid amount exceeded rmb 100 million in a single month, ranking first in the industry.
in addition, according to statistics from phoenix.com technology, in the ai big model project winning bids in august, iflytek ranked first in both the number of winning bids and the amount of winning bids, followed by baidu, alibaba cloud, china telecom, huawei and other companies.
since the beginning of this year, iflytek has been leading the industry in large-scale model bidding projects and has been firmly in the first echelon of large-scale model implementation. currently, important areas of large-scale model bidding, such as central state-owned enterprises, education and science, operators, and finance, are all key business scenarios.
the reason is that, on the one hand, iflytek's spark big model has been upgraded to version v4.0, which is comparable to gpt-4 turbo. while leading in technology, it continues to explore innovations on the scene side, bringing more possibilities to industry customers on the business side. on the other hand, iflytek and huawei have jointly built the first wanka big model computing power base "feixing no. 1", which has the leading advantage of independent control, and is more in line with the needs of central state-owned enterprises for independent and controllable big models and computing power bases.
it is worth noting that large-scale comprehensive model projects with a project amount of more than 10 million yuan are increasing rapidly. such projects have put forward higher requirements on manufacturers' computing power management, ai platform, large models, intelligent application landing and other capabilities.
iflytek spark is leading the central state-owned enterprise market
the central state-owned enterprise market is leading the trend of big model implementation. on the one hand, it has the motivation to embrace big model technology to realize the digital upgrade of enterprises. on the other hand, its open and complex scenarios have also become an important market for big models to realize their value.
today, iflytek spark big model is leading the market of central state-owned enterprises. according to public data, iflytek has become a big model partner of many leading enterprises such as china national petroleum corporation, china national energy group, china green development, china insurance group, bank of communications, chery automobile, china faw, volkswagen, haier group, and midea group, and has joined hands with industry leaders to promote the implementation of big model scenarios.
take the recent china national petroleum kunlun big model as an example. as a giant in the energy field, china national petroleum, together with china mobile, huawei, and iflytek, jointly released the kunlun big model with 33 billion parameters on august 28. this release not only reflects the "strong alliance" between the "national" energy group and the leading enterprise in the field of artificial intelligence, but also the commercialization of the "general big model + professional big model" has taken solid steps. it also confirms that the national general big model of domestic computing power has become the preferred solution for central enterprises to solve the application of big models.
the press conference also demonstrated detailed results of the application of large models for seismic interpretation, logging processing and interpretation, employee assistant, and industry experts. among them, the "employee assistant" has been used by hundreds of thousands of employees of cnpc, and is backed by the iflytek spark domestic large model base and the core technical support provided by iflytek for the research and development of "employee assistant" and "industry experts".
it is reported that the kunlun big model passed the national generative artificial intelligence service registration on august 23. it is the first big model to pass the registration in china's energy and chemical industry. its release marks the release of huge value of big models in the national strategic industry field.
at a recent investor exchange meeting of iflytek, iflytek also released an important message, that is, in the solution for central state-owned enterprises, it will adopt a two-pronged approach of "general large model + special large model" to reduce commercial costs and improve usage efficiency. it is no longer a simple addition effect, but a multiplier effect. looking through the phenomenon to see the essence, the large model strengthens the "emergence of wisdom" with rapid iteration, and relies on industry scenarios to release faster, higher, and stronger productivity, and is becoming a new engine to promote the digital transformation of thousands of industries.
