Last released Jun 2, 2026
2-bit quantization with fused Metal dequant kernels for Apple Silicon — up to 8× faster local LLM inference
Supported by