1、The total industrial output value per unit of land in Shanghai’s Industrial Parks exceeds RMB8.5 billion/square kilometer, and the added value of the industrial chain continues to increase
In Jiading Anting Automobile Town, the factory of the foreign-funded enterprise, Aptiv, is very busy. Benefiting from the rapid development of China's new energy vehicle industry, Aptiv, which is mainly engaged in automotive connectors, has grown rapidly in recent years, and its sales in China increased by 12% in Fiscal Year 2023.
Statistics shows that in 2014, the total industrial output value per unit of land in Shanghai's industrial parks was RMB6.91 billion/square kilometer. In 2024, this figure exceeded RMB8.5 billion/square kilometer. Four batches of 55 characteristic industrial parks show a unit land output of nearly RMB11 billion/square kilometer, and the added value of the industrial chain continues to increase.
The impact of new energy vehicles is not limited to the automobile industry. The upstream demand for new energy vehicles is transmitted to the back end of the industry step by step, and the upgrading of material products is imminent. Last year, Baosteel released BCB EV Pro, a new-generation and upgraded version of ultra-light and high-safety pure electric body-in-white, achieving three major upgrades in safety, materials, and low carbon emissions. The steel battery pack solution for new energy vehicles, which was first developed in China, has also been serialized, and has been applied in more than 10 models.
The development of new energy vehicles requires the participation of industrial robots in manufacturing. The high-end manufacturing industry with high added value is greatly promoting the development of Shanghai's robotics industry. Nowadays, domestic robots have almost evenly divided the market with foreign robots. In the next few years, domestic robots will occupy a larger market share. The confidence for domestic robots to overtake comes from the overall localization of the entire industrial chain. The three core components of industrial robots, namely reducers, servo motors, and controllers, have all been replaced by domestic products.
2、20 domestic large models compete for the champion
Recently, Shanghai Artificial Intelligence Laboratory and Mota Community launched Compass Arena, a Chinese large language model evaluation arena, which for the first time gathered a full lineup of domestic mainstream large models, including Ali Tongyi Qianwen, Baidu Ernie Bot, Tencent Hunyuan, ByteDance Doubao, Shusheng Puyu and other 20 domestic large models competed for the champion of Chinese large models.
At present, various large model lists are emerging one after another. Chatbot Arena, the large model arena launched by Large Model Systems Organization (LMSYS Org), is one of the most credible large model arenas overseas, but the Chinese level of this arena is relatively low and insufficient. The launch of Compass Arena will effectively fill this gap.
Compass Arena, the large language model evaluation arena in China, adopts a blind and open evaluation mode, which can more comprehensively test the strength of the models. Compass Arena has set up random and anonymous battles. Large model contestants become "masked singers". After the model information is hidden, the system randomly matches and conducts competition. Users can freely set questions and use them as judges to make subjective judgments and votes. If a large model accidentally "reports itself", the conversation will be filtered and will not be included in the score. Through thousands of rounds of challenges and user votes, the system will automatically rank the large model.
Compass Arena is jointly built by the Shanghai Artificial Intelligence Laboratory’s OpenCompass Rating System and the Mota Community. The former is responsible for organizing the rating, and the latter is responsible for the introduction of open source models and community creation. Compared with Chatbot Arena, Compass Arena focuses more on Chinese large models, covering all mainstream domestic large models, and can fully evaluate the performance of domestic large models.
