---
license: apache-2.0
base_model:
  - huihui-ai/Huihui-Qwen3.6-35B-A3B-abliterated
  - Qwen/Qwen3.6-35B-A3B
tags:
  - ninfer
  - qwen3.6
  - abliterated
  - rtx3090
library_name: ninfer
---

# Qwen3.6-35B-A3B (Huihui abliterated) — NInfer container v1

Pre-converted **`.ninfer`** artifact for [Don-Chad/ninfer-3090](https://github.com/Don-Chad/ninfer-3090) / NInfer (container **v1**, compatible with prebuilt **v0.3.1**).

| Field | Value |
|---|---|
| Filename | `qwen3_6_35b_a3b_huihui_abliterated.ninfer` |
| Size | ~20.84 GiB |
| SHA-256 | `be263652c8540b9f6d2655fcded3b23eed3193e0c18f9fc29f088a9cd3d6689b` |
| Source BF16 | [huihui-ai/Huihui-Qwen3.6-35B-A3B-abliterated](https://huggingface.co/huihui-ai/Huihui-Qwen3.6-35B-A3B-abliterated) |
| Converter | upstream `tools/convert/qwen3_6_35b_a3b` (same role-based mixed quant as official NInfer 35B-A3B) |

## Why this exists

NInfer loads **only** `.ninfer` at runtime. This upload skips the large BF16 download + local conversion step.

## RTX 3090 note

On 24 GB VRAM use **`--text-only`**. Pin official baselines to container v1 revisions if mixing with neroued artifacts; do not assume HF `main` is v1.

## Quick download

```powershell
hf download ahmed22xa/Qwen3.6-35B-A3B-huihui-abliterated-NInfer `
  qwen3_6_35b_a3b_huihui_abliterated.ninfer `
  --local-dir models
```

## Example CLI

```powershell
.\ninfer.exe models\qwen3_6_35b_a3b_huihui_abliterated.ninfer `
  --prompt "Hello" --max-new 64 `
  --kv-dtype int8 --mtp-draft-tokens 3 --lm-head-draft `
  --no-thinking --text-only
```

## Attribution

Abliteration / BF16: Huihui. Base model: Qwen. NInfer format / converter: Neroued / Don-Chad NInfer projects. See linked model cards and the engine repository for licenses.