EUR/USD
1.15336
-0.074%
Gold
4371.98
-0.415%
Oil
83.033
1.956%
USD/JPY
159.316
0.010%
GBP/USD
1.35006
-0.044%
GBP/JPY
215.091
-0.031%
Real-time News
On August 11th, Behring announced the official open-sourcing of Ling-3.0-tiny, a native hybrid inference model with a total of 7.9B parameters and only 1.3B activation parameters during inference. The open-sourcing release includes three versions: BF16, FP8, and INT4.Iranian Foreign Ministry spokesman Bagaei: If someone raises a sword while trying to talk to you, rest assured, it will not hesitate to strike you.On August 11th, the Shanghai Municipal Commission of Economy and Informatization issued the "Shanghai Software and Information Service Industry Development 15th Five-Year Plan". The plan mentions integrating into the global software and information industry innovation and application ecosystem, vigorously promoting the construction of an open-source system, and leveraging open-source platforms such as the Magnolia Open Source Research Institute, OpenLoong, and OpenDataLab to promote the "first launch" of open-source projects in Shanghai for intelligent software, corpora, algorithms, models, intelligent agents, and humanoid robot hardware and software. It also promotes more flexible open-source benefit distribution mechanisms such as paid open source and dual licenses to attract global talent to participate in open-source solution innovation. The plan supports enterprises in conducting commercial secondary development based on open-source software, innovating the path of "technology open source + commercial application value-added", and cultivating a number of open-source software commercialization enterprises. Finally, it strengthens international open-source cooperation and exchange, supporting enterprises and institutions to participate in the construction of the global open-source ecosystem and the formulation of international rules.On August 11th, the Shanghai Municipal Commission of Economy and Informatization issued the "15th Five-Year Plan for the Development of Shanghais Software and Information Service Industry." The plan emphasizes tackling next-generation model architectures and promoting the exploration of multiple technical routes based on non-Transformer architectures such as state-space models, recurrent neural network variants, and liquid neural networks. It also calls for accelerating the development of cutting-edge basic model technology systems such as physical intelligence, world models, quantum intelligence, and brain-like intelligence. Furthermore, it emphasizes tackling the networking technology of ultra-large-scale intelligent computing power clusters, focusing on core components such as high-performance computing chips (GPUs/NPUs), quantum computing chips (QPUs), high-speed optical interconnects (CPOs), high-bandwidth memory (HBMs), and heterogeneous servers to enhance the supply capacity of intelligent computing hardware and promote the deep integration of domestically developed chips with mainstream large-scale models. Finally, it focuses on new storage and retrieval technologies, digital-model collaboration, and breakthroughs in key technologies such as high-precision heterogeneous processing, native multimodal fusion, and dynamic value alignment to build automated complex reasoning covering the entire lifecycle of corpora.On August 11th, the Shanghai Municipal Commission of Economy and Informatization issued the "15th Five-Year Plan for the Development of Shanghais Software and Information Service Industry." The plan outlines a tiered supply system combining large clusters, small clusters, and edge computing power. It calls for the development of ultra-large-scale intelligent computing clusters (100,000-card level) in Songjiang, Lingang, and Qingpu districts, and thousand-card level intelligent computing clusters in Baoshan, Pudong, and Jiading districts. The plan also guides the transformation of traditional data centers and information communication equipment rooms into hundred-card level edge intelligent computing centers to meet the ultra-low latency computing power needs of enterprises and individuals. Furthermore, it focuses on building Model-as-a-Service (MaaS) platforms for industries such as finance, education, healthcare, culture and tourism, and manufacturing, providing services such as industry application marketplaces, model customization and hosting, agent building, low-code development, API interfaces, computing power provision and management, and AI inference, thereby enhancing the level of intelligent cloud services.
