---

license: apache-2.0

base_model: jbrahy/Qwen3.6-35B-A3B-Never-Never-LAN-uncensored-abliterated

base_model_relation: quantized

pipeline_tag: text-generation

library_name: gguf

language:

• en

tags:

• abliterated
• uncensored
• gguf
• moe
• qwen3
• neverneverlan

---

NeverNeverLAN — Qwen3.6-35B-A3B (abliterated) · GGUF

> ⚠️ Use at your own risk. This model is abliterated (uncensored) — its safety refusals have been removed, so it will generate content the base model would decline, potentially including offensive, harmful, or unlawful material. It is provided as-is, with no warranty of any kind, express or implied. You are solely responsible for how you use it and for compliance with all applicable laws and the base model's license. Do not use it for illegal or harmful purposes.

GGUF quantizations of jbrahy/Qwen3.6-35B-A3B-Never-Never-LAN-uncensored-abliterated — a sparse MoE (~34.7B total / ~3B active). This arch (qwen3_5_moe) was long unsupported by llama.cpp; recent master added it, so these GGUFs now run anywhere llama.cpp does, including Apple Silicon via Metal.

Files

| File | Quant | Size |

|---|---|---|

| NeverNeverLAN-Qwen3.6-35B-A3B-abliterated-Q6_K.gguf | Q6_K | ~27 GB |

| NeverNeverLAN-Qwen3.6-35B-A3B-abliterated-Q4_K_M.gguf | Q4_K_M | ~20 GB |

Requirements

You need a recent llama.cpp — one new enough to include qwen3_5_moe / QWEN35MOE in its arch registry (master, ~b6900+). Older builds will refuse to load these.

Usage

llama-server -m NeverNeverLAN-Qwen3.6-35B-A3B-abliterated-Q4_K_M.gguf \
  --host 0.0.0.0 --port 8080 -ngl 999 -c 16384 --jinja

Because only ~3B params are active per token, decode is fast for a 35B once weights are resident. On a 32 GB Mac, Q4_K_M (~20 GB) fits; raise the Metal wired limit and/or use --cache-type-k/v q8_0 for more context.

Limitations & responsible use

Abliteration removes safety refusals; you are responsible for how you use it. Inherits the base model's biases and knowledge cutoff. Licensed Apache-2.0.