| ---
|
| 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.
|
|
|
