license: apache-2.0
base_model: llmfan46/Qwen3.6-35B-A3B-uncensored-heretic
library_name: basert
pipeline_tag: text-generation
tags:
- basert
- apple-silicon
- quantized
- multimodal
- uncensored
- heretic
Qwen3.6-35B-A3B-uncensored-heretic
BaseRT .base builds of llmfan46/Qwen3.6-35B-A3B-uncensored-heretic for fast local inference on Apple Silicon (Metal).
Qwen3.6-35B-A3B is a hybrid-attention (Gated DeltaNet + periodic full attention) MoE instruct model with a native vision tower; these bundles include the vision encoder. This variant is a community decensor of the original Qwen model via Heretic, converted here unmodified from the bf16 safetensors release — not re-quantized from another quant, so there is no compounded quantization error.
Note: this is an uncensored community finetune. It will comply with requests the original Qwen model refuses. You are responsible for how you use it.
Files
| File | Precision | Size |
|---|---|---|
Qwen3.6-35B-A3B-uncensored-heretic-Q4.base |
4-bit | 20.7 GB |
Qwen3.6-35B-A3B-uncensored-heretic-Q8.base |
8-bit | 36.4 GB |
Usage
curl -LsSf https://basecompute.co/install.sh | sh
basert pull basecompute/Qwen3.6-35B-A3B-uncensored-heretic
basert chat basecompute/Qwen3.6-35B-A3B-uncensored-heretic
Released under the apache-2.0 license, inherited from the base model.