Comment by apitman

I've seen people say the models get more erratic at higher (lower?) quantization levels. What's your experience been?