(opens original source in a new tab)Rest of WorldWhy Silicon Valley is divided over China’s powerful, cheap AI modelsLow-cost Chinese open-weight models are splitting tech executives and Washington over national security and competitiveness. (opens original source in a new tab)Viola Zhou·3 August 2026
(opens original source in a new tab)South China Morning PostWill AI healthcare be China’s new global growth engine?Artificial intelligence (AI) has transformed healthcare for hundreds of millions of people in China. AI-powered apps help them manage their health and seek care, while others help doctors better diagnose and treat their patients. AI analysis of routine scans detects cancers and other conditions physicians might have missed. Personalised cancer vaccines built on AI analysis will soon begin rolling off production lines in China. The country’s leading tech companies have poured resources into... (opens original source in a new tab)Ben Jiang·3 August 2026
(opens original source in a new tab)PandailyMiniMax Desperately Swimming Past the Kill Line: Open-Source H3 Between Kimi K3 Performance Ceiling and DeepSeek Price Floor, With R&D Spending Triple RevenueMiniMax H3 open-source release comes as Kimi K3 raises the performance bar and DeepSeek V4-Flash crushes prices, while MiniMax 2025 financials show 253M R&D against 79M revenue. (opens original source in a new tab)contact@pandaily.com (Pandaily)·3 August 2026
(opens original source in a new tab)PandailyChinese Agent Tops OSWorld With 90.2% Success Rate: InAgent Crosses the 90% Threshold for Computer-Use Agents, Surpassing OpenAI, Google, and Anthropic RecordsInAgent achieves 90.2% task success on OSWorld, the first Computer-Use Agent above 90%, with 100% on system-level tasks, as Harness engineering becomes the new AI competition frontier. (opens original source in a new tab)contact@pandaily.com (Pandaily)·3 August 2026
(opens original source in a new tab)PandailyAlibaba Releases Qwen3.8-Max Official Version: 2.4 Trillion Parameters With 1M Context Window, Long-Run Agent Capability, and Open Source Coming Within a WeekQwen3.8-Max official release delivers 2.4T parameters, 1M context, dramatically improved coding and office capabilities, standing in global first tier with open-source weights promised in a week. (opens original source in a new tab)contact@pandaily.com (Pandaily)·3 August 2026
(opens original source in a new tab)PandailyMiniMax H3 Open-Sourced: H3-Omni-Transformer With 33B Parameters Benchmarks Against Seedance 2.0 at One-Third API Price, Runs Locally on Two RTX 5090 CardsMiniMax releases H3-Base weights on HuggingFace, with 33B dense single-stream Omni-Transformer, 16x spatial and 4x temporal compression VAE, and API pricing at one-third of Seedance 2.0. (opens original source in a new tab)contact@pandaily.com (Pandaily)·3 August 2026
(opens original source in a new tab)South China Morning PostChina’s tech giants race to put AI on delivery riders’ headsChinese e-commerce giant JD.com has launched a smart helmet for food couriers with features enabled by artificial intelligence technology, following similar roll-outs by rivals Alibaba Group Holding and Meituan as they seek to improve rider safety and boost delivery efficiency. The helmet integrates an AI-driven voice assistant with a camera that can “see” the surrounding environment, JD.com’s food delivery unit announced on Monday. The company said the device was designed to minimise... (opens original source in a new tab)Ben Jiang·3 August 2026
(opens original source in a new tab)South China Morning PostChina’s tech giants race to put AI on delivery riders’ headsChinese e-commerce giant JD.com has launched a smart helmet for food couriers with features enabled by artificial intelligence technology, following similar roll-outs by rivals Alibaba Group Holding and Meituan as they seek to improve rider safety and boost delivery efficiency. The helmet integrates an AI-driven voice assistant with a camera that can “see” the surrounding environment, JD.com’s food delivery unit announced on Monday. The company said the device was designed to minimise... (opens original source in a new tab)Ben Jiang·3 August 2026
(opens original source in a new tab)South China Morning PostAlibaba’s AI model Qwen3.8-Max made widely accessible ahead of open-weights releaseAlibaba Group Holding has made its next-generation flagship artificial intelligence (AI) model Qwen3.8-Max widely accessible to global users ahead of an open-weights release next week. The move marks Alibaba’s return to open-sourcing its top-tier AI models after keeping several recent flagship releases proprietary earlier this year. It also signals the firm’s entry into a recent round of powerful releases by Chinese developers aggressively narrowing the gap with leading US labs. The massive... (opens original source in a new tab)Wency Chen·3 August 2026
(opens original source in a new tab)South China Morning PostPotential US ban on Chinese AI models could cost American businesses US$12b a year: reportA potential US ban on Chinese open-weight artificial intelligence (AI) models could cost American businesses up to US$12 billion per year, according to calculations by a US-based academic, as technology firms increasingly turn to cost-efficient Chinese solutions. While the exact economic toll of a ban remains difficult to quantify, usage data from New York-based OpenRouter – a large language model (LLM) aggregator – offers a glimpse into the potential fallout, said Daniel Yue, an assistant... (opens original source in a new tab)Xinmei Shen·3 August 2026
(opens original source in a new tab)PandailyTsinghua Team Open-Sources VeriLoop Coder-E1: Evidence-Governed Spiral Drives Verifiable Recursive Self-Improvement for Repository-Level Code RepairVeriLoop Coder-E1 based on Qwen3.6-27B achieves SWE-bench Verified 85.20, SWE-bench Pro 62.38, Terminal-Bench 2.0 76.40, and DeepSWE 33.63 with narrow-domain PEFT and Self-Harness. (opens original source in a new tab)contact@pandaily.com (Pandaily)·3 August 2026
