MLX

Ollama 0.19 is now powered by Apple’s MLX framework, delivering up to 57% faster prefill speeds and nearly double the decode performance on Apple Silicon. The update also introduces NVFP4 quantization support and smarter caching for coding agents.