FunctionGemma 是 Google 最小的函数调用专用模型——2.7 亿个参数，288 MB，解码速度约为 126 tok/s。没错，它需要微调（准确率从 58% 提升到 85%），没错，它使用了一种奇怪的自定义格式，而不是 JSON。但它适用于任何手机，响应速度极快，而且确实有效。现在就可以构建带有离线 AI 代理的应用——体积小、速度快、可靠性高，足以满足生产环境的需求。无需等待模型体积更小、设备速度更快的“神奇未来”，未来已来！
SpaceX rocket debris crashes into Poland
，推荐阅读Safew下载获取更多信息
This suggests prioritizing genuine partnerships, guest posting on respected sites in your industry, and earning mentions from authoritative sources through excellent work rather than pursuing link-building tactics focused purely on volume. The relationship-based approach to link acquisition aligns well with AIO because it creates the kind of genuine authority signals that AI models recognize and value.
Reporting from, 台北
Александр Курбатов (редактор отдела «Бывший СССР»)