Latest AI News

Amazon announced that starting in 2026, Alexa+ will deeply integrate home services from Angi, allowing users to book hotels, schedule haircuts, and more through natural language conversations without opening an app. AI acts as a universal intermediary, supporting multi-turn dialogue and dynamic demand adjustments, accelerating its transition toward an AI-driven consumer service platform.

Zhihu's 2025 AI Product Rankings, based on user feedback and expert evaluation, offer authoritative market insights. ByteDance's Doubao topped the list as 'Zhihu Users' Favorite of the Year,' highlighting its market leadership.....

Jan introduces Jan-v2-VL-Max, a 30B-parameter multimodal model designed to enhance AI stability in complex automation tasks. Based on Qwen3-VL-30B-A3B-Thinking, it incorporates LoRA-based RLVR technology to improve multi-step operation reliability for long-term task execution.....

Alibaba plans to purchase tens of thousands of AMD MI308 AI chips, designed for the Chinese market and compliant with U.S. export controls, to meet its AI needs. This highlights Alibaba's focus on AI, though the MI308 differs from Nvidia's H20 in cost-effectiveness and security.....

Zhihu's 2025 AI product rankings highlight industry progress, featuring top picks like Doubao, DeepSeek, and Gemini. Doubao leads the user-favorite list, followed by DeepSeek and Tongyi Qianwen. Overseas top three are Gemini, ChatGPT, and Claude, with tools like Cursor also recognized.....

Lima 2.0 shifts focus to AI, introducing a 'sandbox' mechanism that isolates AI coding agents in virtual machines to prevent access to sensitive host files or risky operations, ensuring secure development.....

FaceWall AI secures hundreds of millions in funding to advance edge-efficient large model R&D and ecosystem development, reinforcing its leadership in edge intelligence. As a pioneer in China's edge-side large models, the company has established a technology loop from theory to product, with its MiniCPM series excelling.....

The domestic large model MiniMax opensources M2.1, achieving breakthroughs in multilingual programming, code generation, and tool calling with its 10 billion parameter sparse architecture. It surpasses closed-source flagship models such as Google and Anthropic in authoritative benchmark tests, marking a new stage in the performance of open source coding models.