On September 25th, 2024, the Baidu Cloud Intelligence Conference was held in Beijing. In order to meet the needs of enterprise customers for model invocation, model development and application development, Baidu AI Cloud released Qianfan Model Platform 3.0.
It is reported that in terms of model calling, the upgraded Qianfan platform can not only call nearly a hundred domestic and foreign large models, including the Wenxin series, but also support calling various traditional small models such as voice and visual. While expanding the types of models, Baidu AI Cloud continues to reduce the cost of model invocation. In the past year, the flagship model of Wenxin has seen a price reduction of over 90% and its main models have been fully free, minimizing the cost of innovation and trial and error for enterprises.
Image: Panorama of Wenxin Large Model Family
In terms of model development, for customers who need to customize and fine tune exclusive models, Qianfan 3.0 not only provides a complete large model toolchain, but also supports the development of traditional models such as CV, NLP, and speech, and realizes the unified management and scheduling of data, models, computing power and other resources, providing one-stop large and small model development services for enterprises.
In terms of application development, Qianfan 3.0 has comprehensively upgraded enterprise level retrieval enhanced generation (RAG) from four aspects: retrieval performance, storage expansion, and deployment flexibility, targeting high-frequency application scenarios of enterprise landing large models; For the development of enterprise level agents, Qianfan 3.0 has added support for over 80 official components, including business self orchestration, manual orchestration, knowledge injection, memory capability, and Baidu search.
In addition, the conference also released the intelligent low code application development platform "AI Speed", which enables enterprise level application development through natural language dialogue and can complete application creation with just one sentence.
The continuous improvement of tool platforms has also promoted the explosive growth of the large model industry in the past year. Baidu said that currently, on the Qianfan platform, the daily average number of adjustments for the Wenxin big model exceeds 700 million times. The Qianfan platform has helped users fine tune 30000 big models and developed over 700000 enterprise level applications.
Cui Shengzhi, president of Samsung Electronics in Greater China, said that at the beginning of 2024, Samsung will launch Galaxy AI in the global market and reach a strategic cooperation with Baidu AI Cloud. Based on the Qianfan platform and Baidu Map, it will bring Chinese users the internationally leading localization function of Galaxy AI, including instant search, note taking assistant, transcription assistant, chat assistant, browsing assistant and other mobile AI experiences.
As the big model industry gradually moves towards the deepwater area, AI native applications will also usher in explosive growth. In order to meet the demand of enterprises to directly purchase mature AI applications, Baidu AI Cloud has upgraded three AI native application products for digital people, intelligent customer service, and code efficiency.
Baidu stated that the AI native intelligent customer service application "KeYue", reconstructed based on the Wenxin big model, has achieved significant improvements in understanding complex user intentions and multimodal information communication, making intelligent customer service smarter and more human like. At present, Ke Yue has helped more than 150 million corporate customers with over 500 million interactions.
Based on the capability of large models, the newly upgraded Xiling Digital Human 4.0 supports the rapid generation of 3D digital human images and videos with different makeup styles and industry characteristics based on text. During this conference, Xiling platform announced a significant reduction in the price of 3D hyper realistic digital humans from 10000 yuan to 199 yuan.
The newly upgraded full process AI code efficiency tool "Wenxin Kuai Code" focuses on the business flow of the entire R&D lifecycle, achieving a dual improvement in the efficiency and quality of full process coding development from project takeover to final delivery. At this conference, Wenxin Kuai Code Industry launched two new features: "Enterprise Code Architecture Explanation" and "Enterprise Code Review". Enterprise level code architecture interpretation can achieve intelligent interpretation of engineering architecture in the early stages of project takeover, helping engineers quickly understand business logic; Enterprise level code review can inherit the coding experience of senior engineers and intelligently assist programmers in identifying and filling gaps. At present, Wenxin Kuai Code has served over 10000 corporate clients, helping millions of Chinese developers improve coding efficiency and overall R&D efficiency by 20%.
