SenseTime recently released a new large-scale model system called "SenseNova 4.0", which has achieved comprehensive upgrades in multiple dimensions, aiming to improve the knowledge coverage, reasoning ability, long-text comprehension, numerical reasoning, and code generation capabilities of AI technology. "SenseNova 4.0" has more comprehensive knowledge coverage and more reliable reasoning ability, as well as superior long-text comprehension and numerical reasoning capabilities. In addition, it also supports cross-modal interaction, enabling different forms of media data to be transformed, providing a broader space for innovation in AI applications. SenseChat V4, SenseTime's large language model, has performed exceptionally well in this upgrade. It supports a context window length of up to 128K and achieves a comprehensive evaluation score comparable to GPT-4, surpassing GPT-3.5. This progress not only demonstrates SenseTime's leading technical strength in the field of large-scale models but also provides stronger AI support for industry applications. In terms of practical applications, "SenseNova 4.0" has established deep cooperation with over 500 clients, accelerating the intelligent transformation of various vertical industries. This deep cooperation model allows SenseTime's AI technology to better integrate into industry needs and promote the intelligent process of various industries. To better meet the needs of developers, "SenseNova 4.0" provides various flexible API interfaces and services. These interfaces and services allow developers to easily call various AI capabilities of the large-scale model according to actual application requirements, reducing development costs and improving development efficiency. The improvement of large-scale model performance is based on the enhancement of language modeling capabilities. In this upgrade, SenseTime has upgraded the "SenseChat V4" large language model and supports tokens of different window sizes, further expanding the model's application scope. In evaluations in multiple fields, SenseChat V4 has demonstrated outstanding performance, especially in reasoning and code programming, surpassing GPT-4. In the authoritative test set HumanEval Coding, its pass rate reached 75.6%. In addition, SenseTime has also launched a new multimodal large-scale model called "SenseChat-Vision V4". This model has 30 billion parameters and its image-text understanding capability is at a leading global level. On the authoritative evaluation benchmark test set MME Benchmark, it ranks first with a comprehensive score of 2199.5 (surpassing GPT-4V's 1926.57). This model has been applied in practical scenarios such as intelligent driving, smart cabins, and the power industry, providing strong support for the intelligent upgrade of various industries. At the same time, SenseTime's self-developed large-scale model "SenseMirage V4" has also been upgraded. The model's parameter size has been increased to billions, and its performance in semantic understanding and image quality details has been improved through algorithm optimization. Now, users can generate movie-level posters with just text input, bringing revolutionary changes to the field of creative design. In addition, combining the Adversarial Distillation algorithm, SenseTime has also launched the SenseMirage-Turbo V4 version, which achieves 10 times faster inference speed compared to the basic version. In addition to technological innovations, SenseTime actively promotes the practical application of large-scale models in real-world scenarios. In order to lower the threshold for using large-scale models, SenseTime has introduced a data analysis tool called "Office Little Raccoon". This tool can be used without programming or complex operations. Users only need to input questions in natural language, combined with SenseTime's large-scale model system's intent recognition, logical understanding, and code generation capabilities, to automatically convert data into meaningful analysis results and visualized charts. This product is more suitable for China's data analysis needs, providing industry users with a more convenient and efficient data analysis solution. In summary, the large-scale model system of "SenseNova 4.0" is an important innovative achievement of SenseTime in the field of AI. Through the construction and upgrade of this system, SenseTime continues to promote the advancement of AI technology and the expansion of application scenarios. In the future, SenseTime will continue to advance the construction of the "SenseNova" large-scale model system, achieve more innovative application scenarios, and promote the important role of AI technology in more fields.