Ollama nearly doubles Gemma 4 speed on Apple Silicon, on by default
Ollama's latest build nearly doubles Gemma 4 token generation on Apple Silicon with multi-token prediction. It's automatic, needs no config, and doesn't change the output.
Ollama's latest build nearly doubles Gemma 4 token generation on Apple Silicon with multi-token prediction. It's automatic, needs no config, and doesn't change the output.