You can now convert EXL3 quants on Apple Silicon Mac

EXL3 quantization, previously limited to CUDA environments and high-end RTX cards, is now accessible on Apple Silicon Macs, allowing users with 64GB+ memory to run and convert these models. Notably, the MiniCPM5 and Qwen3.6-27B models show competitive performance with mean KLD metrics comparable to those processed on RTX hardware. This development enhances the accessibility of high-fidelity quantization for practitioners, enabling more efficient deployment of models on consumer-grade hardware.