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！
Head office job losses part of plan for more separation between supermarket and Argos businesses
。关于这个话题，Line官方版本下载提供了深入分析
The entire pipeline executes in a single call stack. No promises are created, no microtask queue scheduling occurs, and no GC pressure from short-lived async machinery. For CPU-bound workloads like parsing, compression, or transformation of in-memory data, this can be significantly faster than the equivalent Web streams code – which would force async boundaries even when every component is synchronous.
Lepora is currently working on a robotics project under the UK government's Aria research and development scheme.