China AI

AIRichard Bian, head of product operations for Ant Ling, says the OpenAI research agents' Hugging Face hack was an inevitable discovery rather than simply a disaster.

AIYuanjie Semiconductor Technology expects third-quarter net profit of 403 million to 453 million yuan, up 579% to 660% from a year earlier, as data-centre demand drives shipments.

AIExperts at a Rest of World event in San Francisco said new open-weight models from American and European companies still trail dominant Chinese systems on performance.

AISupcon Technology signed a strategic partnership with Alibaba's Qwen Work, an enterprise general-purpose agent product, to deploy agents across its marketing chain from lead collection to payment settlement.

AIAlexander Doria says interpretability analyses of modern models overlook data shape as a likely cause of superposition.

AITencent Hunyuan says Hy Memory 2.0 is out, with an open-source model, framework, and MemHub.

Why it matters: The post names Hy Memory 2.0's open-source model, SDK, and MemHub sync tool, showing how an agent memory system would be adopted and tested across AI tools.

AIKling AI has released Kling 4.0, which generates native video up to 30 seconds long in a single generation.

AITencent Hunyuan releases Hy-Memory-30B-A3B, a compact writer model with about 3B active parameters that decides what to retain in external memory frameworks such as Mem0, MemOS, and Hy-Memory.

Why it matters: The model card lays out a two-stage training method and cross-harness benchmark results, making it useful for comparing memory writers against general-purpose LLMs.

AIKling AI says Kling 4.0 is officially released, positioning it as a tool for turning inspiration into new visual worlds.

AIKuaishou's Kling AI says Kling 4.0 is now fully available on klingai.com, with native clips up to 30 seconds and audio included. The company says 4K and 1080p 10-bit HDR output and an enterprise API are coming soon.

Why it matters: The article separates features new to the public rollout from those already previewed in September, showing which capabilities are available now and which remain pending.

AIAgiBot is rolling out its YuanSheng ecosystem program to more than 20 Chinese universities, including Peking University and Tsinghua University, and plans to place over 1,000 robots on campus.

AI4Paradigm and RiseUnion have open-sourced ModelSphere under Apache 2.0, a platform that manages compute, model serving and runtime tuning above engines like vLLM and SGLang.

AITencent's Toast app now lets Pro users import external Android APK files and H5 source code, then edit them in Toast before filing for copyright and submitting to MyApp.

AIKling AI posts a short message with the hashtags #Kling4 and #KlingAI, with no details on features or availability. A background post from @koldo2k says Kling 4.0 is out, but the Kling AI post itself gives no specifics.

AIPromptArmor Threat Intelligence reports that open-weight models, including DeepSeek, Kimi, GLM, and Qwen, were 88% more susceptible to real-world indirect prompt injection cyber attack chains than proprietary models across 9,800 trials.

AIKling AI's post says "Every idea deserves its shot" alongside the #Kling4 and #KlingAI tags, announcing Kling 4.0 and Kling 4.0 Flash.

AIArthur Mensch argues Europe should not leave frontier AI models to the United States and China, calling that idea a fundamental mistake.

AINathan Chen Guangyu, a 17-year-old AI researcher, announced on social media that he has left Moonshot AI after less than a year and is heading to the United States.

AIOpenCode says GLM 5.3 Flash is now available for free in its platform. The post gives no pricing details, usage limits, or further specifications.

AIPrismML releases the Ternary Bonsai 2 series in 2B, 4B and 9B sizes, built on Alibaba's Qwen 3.5 base and using ternary weights of -1, 0 and +1, with storage of about 0.43GB, 0.95GB and 2GB. The company says the models cut tokens by about 40% on standard instruction and daily math tasks for the 9B model, and reaches up to 5x throughput and 3x energy efficiency versus full-precision counterparts on hardware such as the RTX 4090 and M5 Pro.

AIBaidu's autonomous vehicles will carry more than 1,000 Hongkongers in a city-first public trial, according to the South China Morning Post. The post shares a link to the article without further details on routes, duration, or operators.

AIKling AI's Kling 4.0 uses text prompts and reference images or videos to change existing footage, including replacing a product, revising a character, or changing a setting.

AIStepFun has released open weights for Step 5 Preview, a sparse MoE model with 600B total parameters and 27B active per token.

AIAlibaba's Qianwen Office e-commerce edition has over 20% of its paying users from e-commerce, according to CEO Chen Yusen at the October 14 launch.

AIDasha Da, a Chinese drone swarm company, says equipment sales make up about 70% of its revenue while its drone shows gain wide exposure.

AIBMW says the new-generation BMW iX3 will carry a China-specific cockpit AI built with Alibaba on Alibaba's Qwen large model, with Zebra handling cockpit applications and vehicle integration.

AIKling AI released Kling 4.0 on October 15, a video generation model that produces up to 30 seconds in a single native generation.

AIThe Hong Kong Productivity Council opened the city's first experimental embodied AI hub in Kowloon Tong on Thursday, with seven mainland and Hong Kong companies, including Unitree Robotics and Galbot, participating.

AIXiaomi has open-sourced MiDashengLM-Spatial, an roughly 8-billion-parameter audio model under Apache 2.0 that describes sounds and reasons about their direction, distance and movement from binaural input.

AIHuawei, MetaX, Biren and Moore Threads each announced Day-0 support for StepFun's Step 5 Preview, whose weights opened on October 15, meaning the model could be loaded and served on their hardware from release.

AIOpenCode says calls to DeepSeek V4.1 Flash inside its platform no longer use plan credits, so the model is free there.

AIThe Hong Kong Trade Development Council's Autumn Electronics Fair and electronicAsia opened on October 13 at the Hong Kong Convention and Exhibition Centre, running four days with more than 3,200 exhibitors from 15 countries and regions.

AIMathematicians and Shanghai AI Lab's Intern-S2 have resolved Ghys's conjecture in complex dimension three, a problem open for over 30 years. All four main theorems have been formalised and checked in Lean.

AIAnthropic says its latest defenses reduced the reasoning tokens attackers can harvest from Claude by 90–99% over recent months.

AIKuaishou's Kling AI has officially launched its Kling 4.0 video model after a Flash preview for some subscribers in late September.

AIAfter a 30-day public notice period, ByteDance's Doubao Phone Assistant published a list of 155 third-party apps that refuse AI screen automation under its SAEP protocol.

AIAMD and Embedded LLM report that an eight-GPU MI355X system serving Kimi K3 with vLLM delivers 24.4% more total tokens per second per GPU than NVIDIA HGX B300 at 50 tokens/s/user P90 interactivity on the AgentX benchmark.

Why it matters: The post traces optimizations for serving Kimi K3 on AMD MI355X, linking vLLM serving choices to GPU kernels and hardware limits, useful for diagnosing bottlenecks in other inference stacks.

AIComfyUI launches a two-week contest offering $25,000 in prizes for open-source rebuilds of Higgsfield products such as Genjutsu and Cinema Studio.

AIArtificial Analysis reports Vera Rubin delivers about 3.4x to 3.6x GB300's throughput per MW on DeepSeek V4.1 Flash, at roughly 230 to 350 tok/s per user.

AIArtificial Analysis reports that Vera Rubin serves GLM-5.2 at about 3x the throughput per MW of GB300 at roughly 165 tok/s per user.