Ollama MLX Nearly Doubles LLM Speed on Your Mac: 58 to 112 tokens per second
SMRTR summary
Ollama 0.19 nearly doubled LLM speed on Apple Silicon Macs — from 58 to 112 tokens per second — by switching from llama.cpp to Apple's MLX framework. However, most users won't see the improvement unless they download new safetensors-format models, since existing GGUF files still use the old engine.
SMRTR provides this summary for quick context. The original article belongs to GitConnected.
Read the original article