Mixed-precision MLX quantization of huihui-ai/Huihui-Qwen3.6-35B-A3B-abliterated, quantized with MLX Smart Quantize (MSQ) — my own sensitivity-based mixed-precision quantization method for Apple Silicon. It measures per-layer NMSE and assigns optimal bit widths automatically, combining architecture knowledge with measured data.