Not long ago, LMArena, an internationally renowned arena for big models, released its latest ranking, with multiple domestic big models ranking among the top. For example, Baidu Wenxin's large model has the second highest text ability in the world and the first in China, with outstanding performance in creative writing, understanding complex and long problems, and following instructions. Its visual understanding ability ranks first in China and among the top in the world; The Zhipu GLM-4.6 model ranks ahead in terms of coding ability, comparable to multiple international mainstream models. Since the popularity of DeepSeek at the beginning of this year, the international artificial intelligence field has sparked a "China fever", and large models have become a "battleground" for international competition in artificial intelligence. With breakthroughs in multiple dimensions such as technological innovation, practical application, and ecological construction, China's large-scale models are entering the world's top tier and becoming the intelligent foundation leading the leapfrog development of productivity. Intelligence itself is the biggest application, and the speed of technological iteration is the only 'moat'. ”At the 2025 Baidu World Conference held not long ago, Robin Lee, the founder of Baidu, said that in the past year, the ability of big models has gradually stepped out of the category of chat robots. Digital human technology, code agent technology, and even the technology of self evolution in general scenarios to find "global optimal solutions" have made great progress. China is one of the countries with the fastest iteration speed of global big model technology. As of July this year, China has released 1509 large-scale models, ranking first in the world in terms of quantity and accounting for 40% of the total number of large-scale models worldwide. The "2025 Artificial Intelligence Index Report" released by research institutions at Stanford University in the United States shows that the quantity and quality of high-performance AI models in China are constantly improving, and by the end of 2024, the performance of top models at home and abroad will be comparable. In the face of global competition, technological innovation has always been the source of confidence for domestic large-scale models. The latest Wenxin Large Model 5.0, recently released, is a new generation of native full modal large models with a parameter count of 2.4 trillion. It adopts native full modal unified modeling technology and has full modal understanding and generation capabilities, supporting input and output of various information such as text, images, audio, and video. The person in charge of Baidu introduced that the team has made breakthroughs and innovations in the underlying design concept of the model, abandoning the previous idea of "splicing" different information processing modules such as text, visual, auditory, etc., and instead building a single and coherent framework from scratch. Under this framework, the model is endowed with the ability to simultaneously understand and generate diverse heterogeneous information such as text, images, audio, and even videos. This native multimodal design demonstrates unprecedented collaboration and fluency in handling cross domain tasks, laying the foundation for achieving deeper levels of "synesthesia" intelligence. The technological innovation of Chinese large-scale models has also attracted the attention of global developers. A joint report by the Massachusetts Institute of Technology and the open-source community Hugging Face shows that in the past year, the global download share of open-source artificial intelligence models developed in China reached 17.1%, surpassing the United States' 15.8% and ranking first in the world. Not long ago, in the latest ranking released by Hugging Face, Baidu Wenxin Multimodal Thinking Model successively topped the Global Model General Trend List and the Global Multimodal Model Trend List. Currently, Baidu has led over 1000 open source projects and has accumulated over 21000 community contributors, "said Chen Shangyi, Chairman of Baidu's Technical Committee. The competition between large models is not simply a "parameter competition", but a comprehensive competition of underlying hardware, platform ecology, and so on. In terms of underlying computing power, domestic AI computing chips represented by Kunlun Chip, Ascend, Cambrian, etc. are rapidly rising. The latest generation Kunlun chip product released by Baidu has achieved significant improvements in performance and cost, providing Chinese enterprises with more cost-effective AI computing power. At the same time, independent AI development platforms such as Baidu PaddlePaddle, Huawei Shengsi, and Kuangshi Tianyuan also provide complete solutions that are more "hands-on and grounded" for large model landing applications. The value of a large model lies in its application. With its large population and diverse industry application scenarios, China has become the world's largest, most diverse, and rapidly developing "testing ground" for large-scale models. For example, in the field of electricity, which is related to national economy and people's livelihood, relying on the Wenxin big model, Baidu supports State Grid Corporation of China to launch the Guangming Power big model, build a unified open artificial intelligence technology innovation platform covering the company headquarters and 27 provincial branches, and promote unmanned aerial vehicle intelligent inspection, achieving an annual inspection of 5 million towers and reducing the number of manual tower visits by 40%. With the full chain autonomous and controllable capability covering "chip framework model application", Baidu has cooperated with enterprises in different industries to widely apply big model technology in fields such as education, healthcare, finance, and intelligent manufacturing. At present, the number of PaddlePaddle Wenxin ecological developers has reached 23.33 million, and the number of service enterprises has reached 760000. When AI capabilities are internalized and become a native ability, intelligence is no longer a cost, but a productivity. ”Robin Lee said. (New Society)