license: apache-2.0
base_model:
- Qwen/Qwen3.8-27B
pipeline_tag: image-text-to-text
library_name: gguf
language: - en
- zh
tags: - gguf
- abliterated
- qwen
- qwen3
- qwen3.8
- llama.cpp
- uncensored
- ai-red-team
- red-teaming
- vision-language
- mmproj
- mtp
- function-calling
- reasoning
- imatrix
- conversational
Qwen3.8-27B-OrcaRouter-Abliterated-IQ5_KS-GGUF
MTP-capable IQ5_KS quant of OrcaRouter's abliterated Qwen3.8-27B for use with ik_llama.cpp, created using Ubergarm's tensor recipe for Qwen3.6-27B and his Qwen3.8 importance matrix.
The 64 repeating layers follow the same per-tensor recipe as the base Qwen3.8 IQ5_KS quant. The embedded MTP head is retained; its weight tensors are Q8_0 and its norm tensors remain F32.
Files
| File | Purpose | Size | SHA-256 |
|---|---|---|---|
Qwen3.8-27B-OrcaRouter-Abliterated-MTP-IQ5_KS.gguf |
MTP-capable IQ5_KS build | 18.964 GiB | FED8D67851DA7ED867F0F922808AEC83FA64963DF520F9413A6A4B2D4531E80A |
MTP usage
./llama-cli \
-m Qwen3.8-27B-OrcaRouter-Abliterated-MTP-IQ5_KS.gguf \
-ngl 999 \
--spec-type mtp:n_max=1,p_min=0.0
Chat Template
The GGUF preserves the chat-template metadata embedded in OrcaRouter's source checkpoint.
Notes
This repository contains the text-model GGUF only. Use a matching vision projector (mmproj) separately for image input. The source model is abliterated, so its refusal behavior differs from the original Qwen model; use appropriate safeguards for your application.
Sources and tooling
- Abliterated source model: orcarouter/Qwen3.8-27B-Uncensored
- Tensor recipe: ubergarm/Qwen3.6-27B-GGUF
- Importance Matrix: ubergarm/Qwen3.8-27B-GGUF
- Individual-tensor quantizer: ik_llama.cpp