post-training quantization method for GGUF models that uses an imatrix-driven priority queue to maximise theoretical quality per megabyte.