Huihui Qwable 3.6 27b abliterated MTP needs about 23.2 GB of memory at Q4_K_M (a 20.5 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 3 quants, from 20.5 GB (Q4_K_M) to 54.7 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 8.7 GB.
Qwable 3.6 27B – a fine‑tuned Qwen model for structured, code‑centric instruction following
Huihui Qwable 3.6 27B is a 27 billion‑parameter model based on unsloth/Qwen3.6-27B. It has been instruction‑tuned on a cleaned Fable 5 reasoning dataset to produce organized, step‑by‑step responses. The checkpoint is fully fine‑tuned and contains no MTP layers.
Good at
- Structured instruction following
- Code generation and editing
- Technical debugging assistance
Limits
- No MTP layers
- Vision capability present but not fully supported in all runtimes
- Requires large VRAM for full precision
Best for
- Coding assistants
- Technical writing
- Local agent experiments
Who it’s for. Researchers and developers who need a locally deployable, instruction‑tuned Qwen model for coding and technical assistance. The model is suitable for experimentation and local inference in llama.cpp or LM Studio.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 20.5 GB | 23.2 GB | Too big | Get file → | |
| Q4_K_M | 27.8 GB | 30.4 GB | Too big | Get file → | |
| F16 | 54.7 GB | 57.3 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_K_M (20.52 GB file, ~23.2 GB memory at 8 k context). For 16 GB VRAM, Q4_K_M (27.76 GB file, ~30.4 GB memory) is suitable. Full precision F16 requires 54.66 GB file and ~57.3 GB memory, which is beyond typical consumer GPUs.
What GPU do you need for Huihui Qwable 3.6 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: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: the best quant that fits is Q4_K_M (20.5 GB file, about 23.2 GB in use).
- 32 GB VRAM: the best quant that fits is Q4_K_M (27.8 GB file, about 30.4 GB in use).
- 48 GB VRAM: the best quant that fits is Q4_K_M (27.8 GB file, about 30.4 GB in use).
Which Huihui Qwable 3.6 27b abliterated MTP quant should you choose?
Q4_K_M is the usual starting point for Huihui Qwable 3.6 27b abliterated MTP: it balances file size and quality well. Here it is a 20.5 GB file.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Huihui Qwable 3.6 27b abliterated MTP
How much VRAM does Huihui Qwable 3.6 27b abliterated MTP need?
About 23.2 GB at Q4_K_M with an 8K context window. That is the 20.5 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 Huihui Qwable 3.6 27b abliterated MTP on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 23.2 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Huihui Qwable 3.6 27b abliterated MTP GGUF?
It depends on the quant: Q4_K_M is 20.5 GB, Q4_K_M is 27.8 GB, F16 is 54.7 GB.
What context length does Huihui Qwable 3.6 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 Huihui Qwable 3.6 27b abliterated MTP GGUF?
The files are on Hugging Face in the huihui-ai/Huihui-Qwable-3.6-27b-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 Huihui Qwable 3.6 27b abliterated MTP use?
Hugging Face lists the license as mit. 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?
Yes, using the Q4_K_M quantized file (20.52 GB) will fit within 8 GB VRAM with an estimated 23.2 GB memory usage at 8 k context.
Does this model support vision tasks?
It has detected vision abilities, but the checkpoint contains no MTP layers and may not function fully in all vision‑enabled runtimes.
Is the model safe for public deployment?
The model has reduced safety filtering; it is recommended for research or controlled environments only.