Sources indicate that Huawei's Zurich Research Center has introduced a novel open-source quantization technique named SINQ (Sinkhorn-Normalized Quantization). This innovative approach can significantly diminish memory needs while maintaining the output quality of large models intact. Presently, SINQ is accessible as open-source on both GitHub and Hugging Face, licensed under Apache 2.0. This enables enterprises and research organizations to utilize, adapt, and commercially implement it without any cost.