ByteDance and POSTECH have developed a new 1.58-bit quantized FLUX model, significantly reducing memory usage by 7.7 times while enhancing performance. This model addresses the challenges of deploying AI-driven text-to-image generation on resource-limited devices, providing a promising solution for the future of generative AI.

BytePush has developed the 1.58-bit quantized FLUX model, achieving a 7.7-fold reduction in memory usage while enhancing performance. This innovation addresses deployment challenges for AI-driven text-to-image generation on resource-constrained devices.