Kunlun Wey announced today that its TianGong series has made significant progress, with the performance of TianGong 3.0 large-scale model significantly improved, and the public beta of TianGong SkyMusic music large-scale model opened simultaneously. This series of upgrades and openings marks Kunlun Wey's continuous innovation and breakthroughs in the field of artificial intelligence. As Kunlun Wey's star product, TianGong 3.0 has become the world's largest open-source MoE model. The model has an astonishing 400 billion parameters, surpassing the previous 314 billion parameter Grok-1, demonstrating its powerful computing and processing capabilities. In the fields of semantic understanding, logical reasoning, generality, generalization, uncertain knowledge, and learning ability, TianGong 3.0 has shown significant performance improvements. It is worth mentioning that its capabilities in key areas such as mathematics, reasoning, code, and cultural creativity have improved by more than 30%, demonstrating the model's wide range of potential applications in multiple fields. In addition, TianGong 3.0 has added multiple AI capabilities, including multi-round search and comprehensive tool invocation, chart drawing, research mode, enhanced mode, image editing and expansion, etc., providing users with a richer and more convenient functional experience. These new features not only enhance the practicality of TianGong 3.0 but also further consolidate its leading position in the field of artificial intelligence. At the same time, Kunlun Wey also announced the official public beta of TianGong SkyMusic music large-scale model to the whole society. As a new member of the TianGong series, TianGong SkyMusic has shown significant advantages in core areas such as vocal and BGM sound quality, naturalness of vocals, and intelligibility of pronunciation, surpassing the industry-leading Suno V3 in comprehensive performance and achieving the honor of the state-of-the-art model in music large-scale models. TianGong SkyMusic adopts the advanced Sora model architecture in the music audio field, with Large-scale Transformer responsible for composition, learning the contextual dependencies of Music Patches, and achieving controllability of music. At the same time, Diffusion Transformer is responsible for the singing part, using LDM technology to restore Music Patches into high-quality audio. This combination of technologies enables TianGong SkyMusic to support the generation of stereo songs with a duration of up to 80 seconds and a sampling rate of 44100Hz, bringing users an immersive music experience. Kunlun Wey stated that the public beta of TianGong SkyMusic will provide a new platform for music creation and appreciation for users. Users can generate personalized music works through TianGong SkyMusic and explore the infinite possibilities of music. At the same time, this will further promote the development of the music industry and provide more creative inspiration and tool support for musicians. The upgrades and openings of TianGong 3.0 and TianGong SkyMusic are important achievements of Kunlun Wey's continuous innovation in the field of artificial intelligence. In the future, Kunlun Wey will continue to increase its research and development efforts and launch more excellent artificial intelligence products to provide users with smarter and more convenient service experiences.