Improved performance and model support with GGUF

Ollama 0.30 introduces enhanced performance and expanded GGUF model compatibility through integration with llama.cpp, delivering up to 20% faster throughput on NVIDIA hardware and enabling Vulkan support by default for AMD and Intel devices. Users can now run additional model families and fine-tuned variants directly from Hugging Face, with support for tool calling capabilities across coding agents and assistants. The update supplements the existing MLX engine for Apple silicon devices while broadening hardware coverage.