Llamafile, Mozilla’s portable LLM runner, gets GPU support and a rebuilt core

Llamafile 0.10.0 is a major rewrite focused on portability and bundling model weights, reintroducing GPU acceleration (CUDA on Linux, Metal on macOS), adding a terminal UI, server and CLI modes, multimodal and speech support (including Whisper), updating llama.cpp for new model support, simplifying the build system, and shipping several example llamafiles while noting remaining unported features and missing Windows GPU support.