Hypernova-60B Claims Better Performance at 4-Bit Precision

Summary

Hugging Face presents Hypernova-60B-2605, a 59B text-generation model described as a compressed 4-bit model that outperforms its full-precision original. The listing highlights quantization-aware healing but does not provide benchmark or implementation details in the available text.