Mole Threads and 360 Group Announce Strategic Partnership in Digital Security and AI
TapTechNews August 2nd news, according to the tweet of Mole Threads' official account today, during the 12th Internet Security Conference, Mole Threads and 360 Group announced a strategic partnership and signed a strategic cooperation agreement on site.
The cooperation between the two parties aims to focus on the innovative applications in the fields of digital security, information and digital services, and large models, and jointly build a stronger and more reliable artificial intelligence ecosystem to provide advanced technical support and services for all walks of life. The two parties will also work together to create a 360 Zhinao Large Model All-in-One, accelerating the development and application of artificial intelligence technology and injecting strong impetus into the development of the large model industry.
According to the introduction, the 360 Zhinao Large Model All-in-One integrates the 360 Zhinao large model and has functions such as large model management and inference deployment. The full-featured GPU of Mole Threads will provide underlying computing power for it, and ensure the balance between the inference efficiency and cost of the large model with a cost-effective hardware solution. The two parties will also jointly develop and optimize the inference framework, and conduct in-depth customization at the software level, and are committed to bringing a more competitive solution for large model inference.
The 360 large model product has also been adapted to Mole Threads. Based on the Kua E (KUAE) kilo-card intelligent computing cluster of Mole Threads, the 360 Group has successfully completed the distributed training tasks of the 7 billion and 70 billion parameter large models. Mole Threads said that the test results show that the whole process of software and hardware is plug-and-play, the training stability of the cluster reaches 100%, and both the training efficiency and compatibility reach the expected.
According to TapTechNews' previous report, in July this year, Mole Threads announced that its AI flagship product, the Kua E (KUAE) intelligent computing cluster solution, expanded from the current kilo-card level to the ten-thousand-card scale. The Kua E ten-thousand-card intelligent computing cluster of Mole Threads takes the full-featured GPU as the base to create a domestic general-purpose acceleration computing platform that can carry ten-thousand-card scale and has ten-thousand-P-level floating-point computing ability, claiming to be specifically designed for the training of complex large models with trillions of parameters.