Tianqi Zhixin WAIC has released the new-generation universal GPU flagship, Tianqai 300, positioning it against Nvidia Hopper.

robot
Abstract generation in progress
Mars Finance News, July 19: During the 2026 World Artificial Intelligence Conference, Tianshi Zhixin officially launched its new-generation in-house architecture’s first product, the general-purpose GPU flagship Tianqai 300. Its energy efficiency is benchmarked against and exceeds mainstream Hopper architecture solutions. It rebuilds the compute foundation for the three trends of “information intelligence—action intelligence—exploration intelligence.” The product is based on a SIMT architecture, optimized around Attention, MoE, AF/PD separation, and large-scale expansion, balancing compute efficiency, latency, and TCO. In Agent scenarios, the first-token latency drops by about 20%, and Decode communication latency drops by 13%; the Cosmos3 world-model inference performance also exceeds Hopper by 10%. Tianqai 300 is currently ready for large-scale application, and is deeply adapting with China’s cloud providers, server manufacturers, and super-node system integrators, supporting deployment from single machines to clusters. (Wide-angle observation)
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned