Apache License 2.0
Upstream: OBLITERATUS/Qwen3.6-27B-OBLITERATED (base_model: Qwen/Qwen3.6-27B)
Quantization: MLX affine int8 (8.501 bits/weight) via mlx_lm.convert -q --q-bits 8 --q-mode affine
Note: q8 affine used because Ollama 0.24 MLX runner has known bug with mxfp8 on Qwen3.5 architecture