China AI Native Industry Insights – 20260917 – Tencent | Alibaba | MiniMax | more
Today’s digest highlights significant advancements from Tencent, Alibaba, and Qoder Cloud Agents. These companies are fostering an ecosystem focused on enhancing AI agent capabilities and scalability. Tencent has open-sourced BrowserSkill, a CLI tool that allows AI agents to utilize browser tabs more effectively, while Alibaba’s Ant Ling team introduces three quantized versions of Ling-3.0-flash-Fin, aimed at refining AI financial analysis. Meanwhile, Qoder Cloud Agents has launched its Agent-as-a-Service platform with version 1.0, designed for seamless production deployment. These developments indicate a strong push towards open-source solutions and improved AI agent integration within various applications. Discover more in Today’s China AI Native Industry Insights.
1. Tencent open-sources BrowserSkill, a CLI tool letting AI agents borrow your browser tab
Tencent released BrowserSkill, an open-source, MIT-licensed CLI tool that connects AI agents to a user’s actual browser instead of a blank automated instance. The tool lets an agent borrow an existing tab, preserving login sessions, while returning control to the user for captchas and confirmation dialogs. As a CLI rather than an MCP server, it works with any agent capable of running shell commands, including Cursor, Claude Code, Codex, and others, and logs every call it makes. Tab-borrowing permission is controlled through browser settings rather than a command-line flag, and the entire tool runs locally.
Read more: https://github.com/Tencent/BrowserSkill
Video Credit: The original article
2. Alibaba’s Ant Ling team releases three quantized versions of Ling-3.0-flash-Fin
Ant Ling, part of Alibaba’s AI research group, announced three quantized versions of its Ling-3.0-flash-Fin model: FP8, FP4, and INT4. The releases target developers building real-world financial workflows, allowing them to select a version suited to their infrastructure, memory, and efficiency requirements. Ling-3.0-flash-Fin is a variant of the Ling-3.0 model line focused on financial applications. The quantized releases aim to lower deployment costs while preserving model performance.
Read more: https://x.com/i/web/status/2099512582248046639
Video Credit: @AntLingAGI on X
3. MiniMax Highlights Open-Source Ecosystem Growth Around H3 Video Generation Model
MiniMax announced its H3 open-weight video generation model, which supports native stereo audio and multimodal reference control, is gaining rapid open-source community adoption. Collaborators including NVIDIA, FastVideo, Nuva Lab, and Alibaba PAI have released tools such as FastH3, Sol-H3, VDN, and distillation-based LoRAs that speed up H3 inference across hardware like DGX Spark, Apple Silicon, and B300 GPUs. These projects include released weights, training code, and inference code, with workflows also integrated into ComfyUI. MiniMax credited the individual contributors and teams driving these optimizations.
Read more: https://www.minimax.cn/news/minimax-h3-open-source
Video Credit: NotebookLM
4. Qoder Cloud Agents 1.0 launches Agent-as-a-Service platform for production deployment
Qoder has officially released Qoder Cloud Agents 1.0, an Agent-as-a-Service platform designed to help enterprises and individual developers deploy and scale AI agents in production environments. The platform has been in testing since May, supporting hundreds of customers across e-commerce, entertainment, consumer electronics, automotive, and real estate industries, with monthly invocations growing from tens of thousands to tens of millions. QCA 1.0 provides cloud-based harness hosting, multi-agent coordination, enterprise integration with skills and MCP, batch processing, high-concurrency elastic scheduling, and cross-session memory capabilities. Real-world use cases include upgrading e-commerce operational tools and scaling customer service quality inspection from 1% sampling to full coverage.
Read more: https://mp.weixin.qq.com/s?__biz=MzE5ODI2NzI5Nw==&mid=2247492087&idx=1&sn=2dffa75a38ddff69c2fee680c3c580e4&chksm=975fa2f166e554df7d458f9b72391323ae4dad6f3366110f71e921aece082c4b09bfd4ae2e32&scene=0&xtrack=1#rd
Video Credit: NotebookLM
That’s all for today’s China AI Native Industry Insights. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.