Read news on 26.2ms with our app.
Read more in the app
llm.c is now down to 26.2ms/iteration, matching PyTorch