Nace's calibrated quantization: MLX searches layer scales, GGUF uses a calibration-text importance matrix. Shipped only if closer to the original.
-
nace-ai/Qwen3.5-4B-MLX-4bit-KQuant
Image-Text-to-Text • 5B • Updated • 115 -
nace-ai/Qwen3.5-9B-MLX-4bit-KQuant
Image-Text-to-Text • 9B • Updated • 117 • 1 -
nace-ai/Qwen3.6-35B-A3B-MLX-4bit-KQuant
Image-Text-to-Text • 35B • Updated • 37 -
nace-ai/Qwen3.8-27B-MLX-4bit-KQuant
Image-Text-to-Text • 27B • Updated • 40