Users can now spin up multi-person group chats with digital employees like "Qwen Xiao Q" that autonomously parse spreadsheets, draft contracts, and assign project tasks.
New mid-November industry data shows Chinese cloud giants shifting from raw server rental to high-margin token APIs, agent hosting, and industry-specific model fine-tuning.
From desktop productivity agents that build full reports from a single sentence to rack-scale Ascend SuperNodes running DeepSeek inference, Huawei cements its vertical AI stack.
At the 2026 Tencent Global Digital Ecosystem Summit, Tencent Cloud signs regional logistics leader Teleport to automate air-cargo routing and customs workflows across ASEAN.
Every elementary, middle, and high school in China will integrate hands-on AI agent building, prompt engineering, and robotics labs into core STEM curricula.
Alibaba moves to spin off its T-Head (PingTouGe) semiconductor division in a $22B Hong Kong IPO right as its new Zhenwu V900 AI accelerator enters volume deployment across Alibaba Cloud.
China’s largest chipmaker SMIC reported record Q3 revenue of $3.19 billion (+34% YoY) driven by domestic Huawei Ascend AI chip orders, warning that AI silicon shortages will last through 2027.
Beijing-based Moonshot AI has launched Kimi K3.1 Pro, a 2.8-trillion-parameter Mixture-of-Experts (MoE) flagship model supporting a 1-million-token context window for autonomous coding and financial analysis. The release coincides with a landmark Menlo Ventures enterprise study showing Chinese open-weight models now account for 89% of global open-source AI spending, led by Kimi at 61%.
DeepSeek and Huawei have forged a deep hardware-software co-design alliance aimed at breaking Chinese AI labs’ dependence on Nvidia CUDA. DeepSeek has open-sourced Huawei Ascend-native versions of TileLang, DeepGEMM, FlashMLA, and DeepEP, and is planning a massive Inner Mongolia data-center supercluster powered by Huawei Ascend 950DT accelerators.
Baidu has released the pre-training checkpoint and API preview for ERNIE 5.5 Flash, a 280-billion-parameter sparse Mixture-of-Experts (MoE) model with 14 billion active parameters trained on 33.5 trillion tokens. Alongside the model, Baidu debuted ERNIE Evolve, an autonomous research agent designed to execute end-to-end R&D workflows.
Tencent Holdings saw its shares rally 6% after launching Hunyuan Turbo (Hy4) on Tencent Cloud. The flagship multimodal model introduces native computer-use capabilities, long-horizon coding, and visual document comprehension across Tencent’s Yuanbao chatbot and WeChat’s 1.3-billion-user ecosystem.
ByteDance’s Seed AI research division has shipped major upgrades to its Seedance video generation and Seedream visual models, introducing multi-round conversational video editing and native alpha-channel transparency for commercial VFX pipelines. Simultaneously, ByteDance is expanding its Doubao ecosystem into physics-aware world models and autonomous coding agents.
What's next in AI.
The daily briefing on AI news, tools and breakthroughs — from every corner of the world.