Qwen3.8 27B abliterated MTP needs about 19.5 GB of memory at Q4_K_M (a 16.8 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 14 quants, from 0.0 GB to 54.7 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 8.7 GB.
Qwen3.8‑27B‑abliterated: a 27B vision‑language model with refusal removed
Qwen3.8‑27B‑abliterated is a 27‑billion‑parameter dense model that retains the original vision tower and language head. The refusal direction has been removed via a rank‑1 orthogonalization, keeping most of the original capabilities intact. It is available in GGUF format for llama.cpp with MTP support.
Good at
- Vision‑language tasks
- Refusal‑free interactions
- Large context (262,144 tokens)
Limits
- No fine‑tuning or distillation
- Requires MTP‑capable files for self‑speculative decoding
- License: Apache‑2.0
Best for
- Research on refusal mechanics
- Safety testing
- Multimodal inference
Who it’s for. Researchers and developers needing a refusal‑free, vision‑language model for safety studies or multimodal inference. The model is suitable for experimentation in llama.cpp environments that support MTP.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.0 GB | 2.7 GB | Fits | Get file → | |
| IQ2_M | 10.0 GB | 12.7 GB | Too big | Get file → | |
| IQ3_XXS | 11.2 GB | 13.9 GB | Too big | Get file → | |
| IQ2_M | 12.6 GB | 15.3 GB | Too big | Get file → | |
| IQ3_M | 12.8 GB | 15.4 GB | Too big | Get file → | |
| IQ3_M | 14.5 GB | 17.2 GB | Too big | Get file → | |
| IQ4_NL | 16.0 GB | 18.7 GB | Too big | Get file → | |
| Q4_K_M | 16.8 GB | 19.5 GB | Too big | Get file → | |
| IQ4_NL | 17.0 GB | 19.7 GB | Too big | Get file → | |
| Q4_K_M | 17.8 GB | 20.4 GB | Too big | Get file → | |
| Q6_K | 22.4 GB | 25.1 GB | Too big | Get file → | |
| Q6_K | 23.0 GB | 25.7 GB | Too big | Get file → | |
| Q8_0 | 29.0 GB | 31.7 GB | Too big | Get file → | |
| F16 | 54.7 GB | 57.3 GB | Too big | Get file → |
For 8 GB VRAM, use IQ2_M (12.62 GB) or IQ3_M (12.77 GB) which use about 15 GB of RAM. For 12 GB VRAM, IQ4_NL (16.04 GB) or Q4_K_M (16.81 GB) are appropriate, using roughly 19–20 GB of RAM. Higher‑bit quants such as Q6_K or Q8_0 require 25–32 GB of RAM.
What GPU do you need for Qwen3.8 27B abliterated MTP?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 2.7 GB in use).
- 6 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 2.7 GB in use).
- 8 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 2.7 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 2.7 GB in use).
- 16 GB VRAM: the best quant that fits is IQ3_M (12.8 GB file, about 15.4 GB in use).
- 24 GB VRAM: the best quant that fits is Q4_K_M (17.8 GB file, about 20.4 GB in use).
- 32 GB VRAM: the best quant that fits is Q8_0 (29.0 GB file, about 31.7 GB in use).
- 48 GB VRAM: the best quant that fits is Q8_0 (29.0 GB file, about 31.7 GB in use).
Which Qwen3.8 27B abliterated MTP quant should you choose?
Q4_K_M is the usual starting point for Qwen3.8 27B abliterated MTP: it balances file size and quality well. Here it is a 16.8 GB file.
The Q8_0 file is 29.0 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Qwen3.8 27B abliterated MTP
How much VRAM does Qwen3.8 27B abliterated MTP need?
About 19.5 GB at Q4_K_M with an 8K context window. That is the 16.8 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Qwen3.8 27B abliterated MTP on an 8 GB GPU?
Yes, with the smaller files: the 0.0 GB file needs about 2.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Qwen3.8 27B abliterated MTP GGUF?
It depends on the quant: file is 0.0 GB, IQ2_M is 10.0 GB, IQ3_XXS is 11.2 GB, IQ2_M is 12.6 GB, IQ3_M is 12.8 GB, IQ3_M is 14.5 GB and 8 more.
What context length does Qwen3.8 27B abliterated MTP support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 2.2 GB at 8K and 8.7 GB at 32K.
Where can I download Qwen3.8 27B abliterated MTP GGUF?
The files are on Hugging Face in the hotdogs/Qwen3.8-27B-abliterated-MTP-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Qwen3.8 27B abliterated MTP use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Qwen3.8‑27B‑abliterated on 8 GB VRAM?
Use the IQ2_M or IQ3_M quant, which estimate 15 GB of RAM usage at 8 k context. 8 GB VRAM is insufficient for full inference.
Does the model support vision tasks?
Yes, the vision tower is unchanged and fully functional in the GGUF build.
What is the license of this model?
The model is released under the Apache‑2.0 license, inherited from the base model.