Views
No views yet
We found that standard quantization significantly degrades quality and can break the model's reasoning chains. For the best results, we highly recommend running it in BF16 or FP8.
