China AI Native Industry Insights – 20260526 – Huawei | OpenBMB | Tencent | more
Discover Huawei’s Tau Scaling Law, BitCPM-CANN, ima AI Copilot, GLM-5.1 API. Discover more in Today’s China AI Native Industry Insights.
1. Huawei introduces Tau Scaling Law to replace Moore’s Law for semiconductor development
At the 2026 IEEE International Symposium on Circuits and Systems in Shanghai, Huawei’s He Tingbo presented the Tau Scaling Law as a new principle for guiding semiconductor industry development. The law proposes replacing geometric scaling with time scaling as the guiding principle for semiconductor and electronic system evolution. Based on this principle, Huawei developed technologies like LogicFolding to compress signal propagation delay and improve transistor density. Huawei projects that its high-end chips could reach transistor density equivalent to 1.4nm process by 2031.
Read more: https://www.huawei.com/en/news/2026/5/ieee-iscas-tau-scaling
Video Credit: @Huawei on X
2. MiniMax and Tsinghua University open-source BitCPM-CANN, China’s first 1.58-bit edge model trained on Huawei Ascend NPU
MiniMax, in collaboration with Tsinghua University and OpenBMB, has officially released and open-sourced BitCPM-CANN, China’s first 1.58-bit ternary large language model trained end-to-end on Huawei’s Ascend NPU platform. The model family includes four sizes (0.5B, 1B, 3B, and 8B parameters) and achieves 90-97.2% performance retention compared to full-precision MiniCPM4 models while delivering approximately 6x memory savings during inference. Using quantization-aware training (QAT) rather than post-training quantization, BitCPM-CANN demonstrates that domestic NPU platforms can handle advanced low-bit training workloads, potentially enabling 60B parameter models to run on mobile devices. All model weights are now available on HuggingFace and ModelScope for developers to download and use.
Read more: https://mp.weixin.qq.com/s/zN3NaWlgWbuXgKjYEtuegQ
Video Credit: @OpenBMB on X
3. ima Opens AI Copilot Feature to All Users and Launches Skill Sharing on Knowledge Platform
ima announced it is removing the waitlist for its Copilot feature, making it fully accessible to all users after more than 100,000 people queued for access. The Copilot is an AI agent with memory and knowledge base integration that can perform tasks across the platform, including reading current pages, searching the web, and executing complex workflows. Additionally, ima introduced a new Skill publishing feature for its knowledge platform, allowing users to package and share their workflows as reusable Skills, transforming the platform from a content-sharing space into a capability-sharing ecosystem. The update also includes official Skills for integrations like WeChat Reading and Tencent Recruitment.
Read more: https://mp.weixin.qq.com/s/R22ySmNraCJ4W4o2nbJu1w
Video Credit: NotebookLM
4. Zhipu AI releases GLM-5.1-highspeed API with 400 tokens per second output speed
Zhipu AI announced the release of GLM-5.1-highspeed API to select enterprise customers, achieving an output speed of 400 tokens per second. The company claims this sets a new speed record for global large language model API services. Unlike typical high-speed models that sacrifice capability for speed, GLM-5.1-highspeed maintains flagship-level performance while delivering ultra-low latency, making it suitable for speed-sensitive scenarios such as AI coding, real-time interaction, and business decision-making. The performance improvements are powered by the TileRT inference engine, which uses system-level optimizations across inference engine, scheduling system, and infrastructure layers.
Read more: https://mp.weixin.qq.com/s/FJn5athj8G4y2narTzMSyA
Video Credit: Hyperframes
That’s all for today’s China AI Native Industry Insights. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.