China Executive Insights
WUWENAI founder Liu Shengxiang bets that embodied AI's bottleneck is data. The startup built a closed-loop Real-to-Sim-to-Real pipeline that he says arrives months ahead of World Labs' SceniX move.
After a 100-million-yuan seed-plus-angel round, Awomo founder Yu Kaicheng makes his first public case for an implicit, concept-space world model that beats the data-driven scaling law he once helped write.
Jinhua and CXMT both started in 2016 to break the Samsung-SK Hynix-Micron DRAM hold. A US blacklist froze one on the eve of mass production, while the other just rang the bell at the A-share IPO.
LightCounting ranks seven Chinese vendors in the global optical-module top ten. As 1.6T ramps, the focus is shifting from pluggable modules to optical transport, silicon photonics, InP substrates, and CPO.
DeepSeek V4 Flash tops OpenRouter's weekly global token-usage ranking at 7.1 trillion tokens, with four of the top five slots now held by Chinese models, forcing OpenAI to cut GPT-5.6 Luna pricing by 80%.
DeepSeek V4 Flash tops OpenRouter's weekly global token-usage ranking at 7.1 trillion tokens, with four of the top five slots now held by Chinese models, forcing OpenAI to cut GPT-5.6 Luna pricing by 80%.
DeepSeek V4 Flash tops OpenRouter's weekly global token-usage ranking at 7.1 trillion tokens, with four of the top five slots now held by Chinese models, forcing OpenAI to cut GPT-5.6 Luna pricing by 80%.
Tencent rolls out its flagship Hy3 LLM internationally via WorkBuddy, Tencent Design Miora, and TokenHub — free on WorkBuddy until Aug 31, 2026.
Tencent has integrated its Hy3 large language model into the international edition of WorkBuddy, marking the model’s first rollout in the global version of the AI workspace. Users worldwide can access Hy3 at no cost through Aug. 31, 2026, Pacific Time, under a limited-time offer. WorkBuddy lets users issue natural-language instructions for tasks including research, data analysis, document and presentation creation, and workplace communications. Its international version launched in May, while Hy3 was released on July 6. Hy3 has 295 billion total parameters, 21 billion active parameters and a context window of up to 256K tokens. Tencent said the model recorded more than 68 times the API calls of its previous-generation model and ranked first on OpenRouter’s global LLM usage leaderboard within a week of launch. [TechNode reporting] Every Wednesday and Friday, TechNode’s Briefing newsletter delivers a roundup of the most important news in China tech, straight to your inbox. Your support helps TechNode continue to provide credible, on-the-ground journalism and industry insights about the Chinese tech industry.
Xiaomi has open-sourced its embodied-AI foundation model Xiaomi-Robotics-1, the company’s technology account announced on Aug. 5. The release covers the full process from real-robot post-training to model deployment and includes code for related benchmark evaluations. Xiaomi-Robotics-1 was pretrained on more than 100,000 hours of UMI data and post-trained on more than 10,000 hours of cross-embodiment data. Xiaomi first introduced the model in July as an “out-of-the-box” embodied-AI foundation model. The open-source release includes links to the project website, GitHub repository and Hugging Face page. [IT Home, in Chinese] Every Wednesday and Friday, TechNode’s Briefing newsletter delivers a roundup of the most important news in China tech, straight to your inbox. Your support helps TechNode continue to provide credible, on-the-ground journalism and industry insights about the Chinese tech industry.
An independent SaferAI audit of Zhipu's GLM-5.2 found cyber-offense capability on par with GPT-5.5 but a complete absence of content-filtering guardrails, exposing structural risks for open-weight releases.
At the 5th CCF Quantum Computing Conference in Shenzhen, six CAS and CAE academicians outlined a national road map for AI-enabled quantum error correction and a unified quantum-supercomputing-AI infrastructure.