license: apache-2.0
base_model: DavidAU/Qwen3.6-27B-F451-AND-TRI-Polar-Ultra-Pro-Writer-Uncensored-Heretic
pipeline_tag: text-generation
library_name: mlx
tags:
- mlx
- optiq
- mixed-precision
- qwen3.6
- uncensored
- creative writing
Qwen3.6-27B-F451-AND-TRI-Polar-Ultra-Pro-Writer-Uncensored-Heretic-OptiQ-4bit
An OptiQ mixed-precision quantization of
DavidAU/Qwen3.6-27B-F451-AND-TRI-Polar-Ultra-Pro-Writer-Uncensored-Heretic,
built for Apple Silicon with MLX.
OptiQ assigns a per-layer bit-width (4-bit or 8-bit here) instead of quantizing
every layer the same, so the layers that need the precision keep it.
- Method: OptiQ data-driven mixed precision (
optiq_mixed_precision) - Average: ~5.6 bits per weight (220 layers at 8-bit, 276 at 4-bit)
- Architecture:
qwen3_5(Qwen3.6-27B)
The per-layer allocation is reused from
mlx-community/Qwen3.6-27B-OptiQ-4bit,
which shares the same Qwen3.6-27B architecture, so no separate sensitivity pass
was run for this merge. This is a text-only quant of the language tower. The
tokenizer is copied from the base model unchanged.
Use
pip install mlx-optiq
optiq serve --model mlx-community/Qwen3.6-27B-F451-AND-TRI-Polar-Ultra-Pro-Writer-Uncensored-Heretic-OptiQ-4bit
Or load it directly with mlx_lm:
from mlx_lm import load, generate
model, tokenizer = load("mlx-community/Qwen3.6-27B-F451-AND-TRI-Polar-Ultra-Pro-Writer-Uncensored-Heretic-OptiQ-4bit")
See mlx-optiq.com for the CLI, the local Lab UI, and the
OptiQ Code agent. The base model and its behavior are DavidAU's; this repo only
changes the quantization.