1.58 Bit FLUX: Top AI raw graph model quantization that runs smoothly even on cell phones

Byte and POSTECH collaborated to launch 1.58-bit FLUX, which compresses AI drawing models by 7.7 times for easy mobile operation; adopts 1.58-bit quantization technology to dramatically reduce model storage, inference memory and increase inference speed; after testing, the compressed model has almost no difference in image quality, supporting high-quality AI image generation.

Search