Qwen3.8-27B-RANA-abliterated needs about 19.2 GB of memory at Q4_K_M (a 16.5 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 19 quants, from 0.0 GB to 53.8 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 8.6 GB.
Qwen3.8-27B-RANA-abliterated – a 27 B parameter vision‑enabled LLM with a refusal‑direction ablation
Qwen3.8-27B-RANA-abliterated is a 27 B parameter model from the Qwen family, built with the qwen35 architecture. It includes a vision tower and MTP speculative‑decoding head. The model is released under the Apache‑2.0 license and is intended for research on refusal‑behavior removal.
Good at
- Vision‑enabled inference
- MTP speculative decoding
- Large context up to 262 144 tokens
Limits
- No safety alignment – use only for research
- Requires 8 k context for memory estimates
- Large VRAM needed for full precision
Best for
- Research on refusal‑behavior removal
- Vision‑based multimodal tasks
- Testing speculative decoding
Who it’s for. Researchers and developers who need a vision‑enabled, large‑context LLM for experimentation with refusal‑ablation techniques. The model is not intended for production use due to its safety‑alignment removal.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.0 GB | 2.7 GB | Fits | Get file → | |
| Q8_0 | 3.2 GB | 5.8 GB | Fits | Get file → | |
| BF16 | 5.9 GB | 8.6 GB | Too big | Get file → | |
| IQ2_M | 10.0 GB | 12.7 GB | Too big | Get file → | |
| Q2_K | 10.7 GB | 13.4 GB | Too big | Get file → | |
| IQ3_XXS | 11.2 GB | 13.8 GB | Too big | Get file → | |
| IQ3_M | 12.6 GB | 15.2 GB | Too big | Get file → | |
| Q3_K_M | 13.3 GB | 15.9 GB | Too big | Get file → | |
| Q3_K_L | 14.3 GB | 17.0 GB | Too big | Get file → | |
| IQ4_XS | 15.1 GB | 17.7 GB | Too big | Get file → | |
| Q4_0 | 15.5 GB | 18.2 GB | Too big | Get file → | |
| Q4_K_S | 15.6 GB | 18.2 GB | Too big | Get file → | |
| IQ4_NL | 15.8 GB | 18.4 GB | Too big | Get file → | |
| Q4_K_M | 16.5 GB | 19.2 GB | Too big | Get file → | |
| Q5_K_S | 18.7 GB | 21.3 GB | Too big | Get file → | |
| Q5_K_M | 19.2 GB | 21.9 GB | Too big | Get file → | |
| Q6_K | 22.1 GB | 24.7 GB | Too big | Get file → | |
| Q8_0 | 28.6 GB | 31.2 GB | Too big | Get file → | |
| BF16 | 53.8 GB (2 parts) | 56.5 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_K_M (file size 16.55 GB, ~19.2 GB memory at 8 k context). For 12 GB VRAM, Q5_K_S (18.68 GB, ~21.3 GB memory) is suitable. For 24 GB VRAM, BF16 (5.95 GB, ~8.6 GB memory) provides the best precision. Memory estimates are approximate.
What GPU do you need for Qwen3.8-27B-RANA-abliterated?
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 Q8_0 (3.2 GB file, about 5.8 GB in use).
- 8 GB VRAM: the best quant that fits is Q8_0 (3.2 GB file, about 5.8 GB in use).
- 12 GB VRAM: the best quant that fits is BF16 (5.9 GB file, about 8.6 GB in use).
- 16 GB VRAM: the best quant that fits is Q3_K_M (13.3 GB file, about 15.9 GB in use).
- 24 GB VRAM: the best quant that fits is Q5_K_M (19.2 GB file, about 21.9 GB in use).
- 32 GB VRAM: the best quant that fits is Q8_0 (28.6 GB file, about 31.2 GB in use).
- 48 GB VRAM: the best quant that fits is Q8_0 (28.6 GB file, about 31.2 GB in use).
Which Qwen3.8-27B-RANA-abliterated quant should you choose?
Q4_K_M is the usual starting point for Qwen3.8-27B-RANA-abliterated: it balances file size and quality well. Here it is a 16.5 GB file.
The Q8_0 file is 3.2 GB, about 0.2 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-RANA-abliterated
How much VRAM does Qwen3.8-27B-RANA-abliterated need?
About 19.2 GB at Q4_K_M with an 8K context window. That is the 16.5 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-RANA-abliterated on an 8 GB GPU?
Yes, with Q8_0: the 3.2 GB file needs about 5.8 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Qwen3.8-27B-RANA-abliterated GGUF?
It depends on the quant: file is 0.0 GB, Q8_0 is 3.2 GB, BF16 is 5.9 GB, IQ2_M is 10.0 GB, Q2_K is 10.7 GB, IQ3_XXS is 11.2 GB and 13 more.
What context length does Qwen3.8-27B-RANA-abliterated support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 2.1 GB at 8K and 8.6 GB at 32K.
Where can I download Qwen3.8-27B-RANA-abliterated GGUF?
The files are on Hugging Face in the abliterant/Qwen3.8-27B-RANA-abliterated-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-RANA-abliterated 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 this model on 8 GB VRAM?
Use the Q4_K_M quantization; it requires about 19.2 GB memory at 8 k context, so 8 GB VRAM is insufficient. A smaller quant such as Q4_K_S would still exceed 8 GB.
Does this model support vision input?
Yes, the vision tower is included and identical to the base model. Vision files are automatically loaded by compatible apps.
Is the model safe for production use?
No. The model is a safety‑alignment‑removed research model. Use only in controlled research environments.