Tag
A 1-bit quantized version of the Hy3 295B model achieves 2.2x faster inference speed compared to the cloud API with no quality loss.