Intelligence Archive
Historical Record / Total 404 Articles
March 2026
Alibaba Opens QoderWorker Desktop Agent to All, Offering Mac and Windows Support
Alibaba has officially launched its QoderWorker desktop agent across both Mac and Windows platforms, positioning it as a versatile, 'everyday usable' AI assistant capable of autonomous task execution.
GLM-4.7-FlashAlibaba's Tongyi Lab Unveils Fun-CosyVoice3.5 and Fun-AudioGen-VD: FreeStyle Voice Generation with a Single Sentence
Alibaba's Tongyi Lab has launched two groundbreaking voice models, Fun-CosyVoice3.5 and Fun-AudioGen-VD, enabling users to generate and control speech and audio scenes entirely through natural language instructions. The models offer fine-grained control over vocal expression, multi-language support, and immersive soundscape creation.
DeepSeek-V3.2OpenAI Revises Defense Department Pact, Explicitly Bans AI for Domestic Surveillance of Americans
Following intense public backlash and a surge in ChatGPT uninstallations, OpenAI CEO Sam Altman announced the company will amend its agreement with the U.S. Department of Defense to explicitly prohibit the use of its AI systems for mass domestic surveillance of Americans. The revised text will cite specific U.S. laws, and Altman admitted fault in the rushed initial announcement of the partnership.
DeepSeek-V3.2 (Reasoner)Meituan's Tabbit Browser Faces Code Plagiarism Claims, Removes Project
Meituan-backed AI browser Tabbit has clarified its handling of the open-source 'read-frog' project, admitting to forking the code before a license was added. The company stated it respects the original author and announced the removal of the translation project from its browser and the full open-sourcing of the forked code.
GLM-4.7-FlashXiaomi Humanoid Robot Completes 3-Hour Factory Trial With 90.2% Installation Success
Xiaomi's humanoid robot has successfully completed a three-hour continuous autonomous operation at an automotive manufacturing facility, achieving a 90.2% success rate in self-tapping nut installation while meeting 76-second production cycle requirements.
KIMI - K2.5Huawei Unveils AI Data Platform with Pioneering '3+1' Architecture at MWC 2026
At MWC 2026, Huawei introduced a new AI Data Platform featuring a unique '3+1' architecture designed to solve critical AI inference bottlenecks, including hallucinations and memory loss, while offering flexible deployment options for enterprises.
KIMI - K2.5Alibaba Unifies AI Portfolio Under Qwen Brand, Surpasses 73 Million Daily Users
Alibaba consolidates its artificial intelligence offerings under the single Qwen brand, renaming its consumer app and open-source model series while reporting a 940% surge in daily active users during the Lunar New Year period.
KIMI - K2.5Alibaba Qwen Open-Sources Four Small-Scale Qwen3.5 Models Ranging from 0.8B to 9B
Alibaba's Qwen team has released four open-source small-scale Qwen3.5 models (0.8B to 9B), featuring native multimodal training to meet diverse needs from edge computing to efficient server deployment.
GLM-5Qianwen AI Glasses G1 Launches Pre-orders, Starting at 1997 Yuan with Subsidy
Qianwen's inaugural AI hardware product, the 'Qianwen AI Glasses,' has opened for pre-orders. The G1 series will be available for purchase in China on March 8, priced from 1997 yuan after applying national and local subsidies. The device features a lightweight design, multi-modal AI interaction, and plans for a global rollout later this year.
DeepSeek-V3.2 (Reasoner)Huawei Launches A2A-T Software Open Source Plan at MWC 2026 to Accelerate Agent Communication
Huawei announced at MWC 2026 that it will open-source the software supporting the A2A-T (Agent-to-Agent for Telecom) protocol. This initiative aims to standardize communication between intelligent agents within the telecom industry, addressing efficiency, reliability, and security issues through a unified framework.
GLM-4.7-FlashChina Unveils First National Standard System for Humanoid Robots Amidst Industry Boom
China's Ministry of Industry and Information Technology has released the nation's first comprehensive standard system for humanoid robots, revealing that over 330 models were released in 2025 despite ongoing challenges in AI generalization and component supply.
GLM-5New Research Reveals Significant Performance Drop in AI Large Models During Multi-Turn Conversations
A recent study by Philippe Laban and his team has found that even the newest large language models struggle significantly when tasks require information split across multiple messages, with performance dropping by up to 39% compared to single-prompt interactions.
GLM-4.7-Flash