So often means q5_0 or q5_1 .

It features 24 audio layers and 24 text layers , providing a significant jump in complexity from the "Small" or "Base" models. Performance vs. Accuracy: The Medium Trade-off