# China AI
Latest news and articles about China AI
Total: 35 articles found
Thinking with Coordinates: DeepSeek’s Move Toward ‘System 2’ Multimodal Intelligence
DeepSeek has released a technical framework that enables AI models to use spatial coordinates as 'visual primitives' in their reasoning process. This innovation bridges the referential gap in multimodal AI, allowing for more precise visual reasoning and industry-leading token efficiency.
Desert Silicon: China’s State-Led Gambit to Power the AI Age
China is leveraging its vast Northwest deserts to create an integrated corridor of green energy and AI computing power. Led by state-owned enterprises, this strategic shift aims to provide the low-cost, sustainable infrastructure necessary for the country's long-term leadership in artificial intelligence.
National Team Play: China’s State-Backed Supercomputing Network Offers Free Access to DeepSeek-V4
China’s National Supercomputing Infrastructure has launched free access to the DeepSeek-V4 model, emphasizing its 1-million-token context capacity. This move integrates state-led computing resources with cutting-edge private AI development to accelerate domestic innovation.
The High Price of Purity: DeepSeek’s Talent Exodus and China’s AI Wage War
DeepSeek's latest technical report reveals a significant loss of core researchers to Chinese tech giants, highlighting the intense competition for AI talent. Despite its technical success, the firm's refusal to accept external funding limits its ability to compete with the massive salary and equity packages offered by VC-backed rivals and Big Tech.
The Agent Pivot: Why 2026 Marks China’s Shift from Large Models to Autonomous Systems
China's AI sector has pivoted from developing foundational models to deploying autonomous 'Agents' in 2026, driven by an explosion in token consumption and a new focus on 'Harness Engineering.' Major players like Alibaba and DeepSeek are now prioritizing task-oriented autonomy and domain-specific expertise to transform AI from a simple tool into a functional 'digital employee.'
JD and Qianxun Tie Up to Bring ‘Embodied Intelligence’ Into Retail at Scale
Qianxun Intelligent and JD Group have signed a strategic cooperation agreement running from 2026 to 2029 to commercialize embodied-intelligence technologies in retail and consumer products. The partnership aims to combine Qianxun’s device and spatial-AI expertise with JD’s logistics and retail reach to accelerate real-world deployments.
The Token Wars Begin: Nvidia’s Vera Rubin vs China’s Low‑Cost Inference Push
At GTC 2026 Nvidia declared the AI era has shifted from training models to continuously generating tokens and presented Vera Rubin, a full‑stack platform it says can cut token costs dramatically. At the same time, Chinese large‑model providers are already undercutting foreign counterparts on token prices and capturing high API volumes, creating a global contest over who will set token pricing and infrastructure standards.
Jieyue Xingchen Launches StepClaw — 50,000 One‑Click Cloud AI Assistant Deployments with a 50M‑Token Trial
Jieyue Xingchen launched StepClaw, a cloud AI assistant platform built on OpenClaw that offers 50,000 one‑click deployment slots with a one‑month free trial including 50 million tokens, compute and storage. The move lowers technical barriers to running assistants, accelerates experimentation, and signals intensified competition over hosted model services in China amid regulatory and moderation challenges.
Tencent’s Secret WeChat AI Agent: Preparing Autonomous Assistants for 1.4 Billion Users — and a Cloud‑Compute Bonanza
Tencent is developing a secret AI agent for WeChat meant to autonomously use mini‑programs to handle tasks for users, with limited trials planned for mid‑2026 and a possible full rollout by Q3. The project mixes in‑house and third‑party models, seeks to monetise heavy cloud usage generated by autonomous agents, and confronts significant privacy, security and regulatory hurdles.
China’s ‘Lobster’ Craze: OpenClaw Agents Promise New Productivity — and New Risks
OpenClaw agents, nicknamed “lobsters,” are spurring a wave of desktop automation in China that promises increased productivity and new business models but also raises steep costs and security concerns. A NetEase salon on March 13 convened industry leaders to share deployment guides, case studies and safety practices as the technology moves from hobby to enterprise adoption.
‘Lobster’ Mania: Cloud and Model Firms Cash In as OpenClaw Sparks a Token Surge
OpenClaw, an open-source AI agent platform dubbed “Lobster,” has ignited widespread deployment in China, prompting cloud vendors and model providers to capitalise on surging token consumption. While the rush is boosting short-term revenues and stock prices, security vulnerabilities, high running costs and a shortage of mature use cases temper the enthusiasm.
Alibaba Approves Resignation of Qwen Lead — A Test for China’s Open‑Model Experiment
Alibaba has approved the resignation of Lin Junyang, a central technical figure behind the open‑source Qwen models, and placed foundation‑model oversight with senior management. The move reassures stakeholders on policy but raises developer fears that Qwen’s open, high‑velocity culture could change, illustrating the friction between engineering ideals and corporate priorities in AI.