Llama 3 8B Instruct abliterated needs about 10.1 GB of memory at Q8_0 (a 8.5 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 3 quants, from 4.9 GB (Q4_K) to 16.1 GB.
Llama 3 8B Instruct abliterated – orthogonalized refusal direction
Llama 3 8B Instruct abliterated is a 8B parameter model with orthogonalized bfloat16 weights. It is tuned to reduce refusal responses while maintaining original instruction capabilities. The model is available in several GGUF quantizations.
Good at
- Reduced refusal responses
- Maintains instruction following
- Compatible with llama.cpp
Limits
- No vision support
- License not specified
- Context limited to 8192 tokens
Best for
- Chatbot applications needing lower refusal
- Instruction‑based tasks
- Testing orthogonalization effects
Who it’s for. Developers and researchers interested in exploring refusal mitigation techniques in LLMs. Useful for those who want to experiment with orthogonalized models without extensive safety tuning.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K | 4.9 GB | 6.5 GB | Fits | Get file → | |
| Q8_0 | 8.5 GB | 10.1 GB | Too big | Get file → | |
| — | 16.1 GB | 17.6 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_K (file size 4.92 GB, estimated 6.5 GB memory at 8k context). For 12 GB VRAM, Q8_0 (8.54 GB, 10.1 GB memory) is suitable. The unknown quant (16.07 GB, 17.6 GB memory) requires larger memory budgets.
What GPU do you need for Llama 3 8B Instruct 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: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: the best quant that fits is Q4_K (4.9 GB file, about 6.5 GB in use).
- 12 GB VRAM: the best quant that fits is Q8_0 (8.5 GB file, about 10.1 GB in use).
- 16 GB VRAM: the best quant that fits is Q8_0 (8.5 GB file, about 10.1 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (16.1 GB file, about 17.6 GB in use).
Which Llama 3 8B Instruct abliterated quant should you choose?
Q8_0 (8.5 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
If memory is tight, the smallest file is Q4_K at 4.9 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Frequently asked questions about Llama 3 8B Instruct abliterated
How much VRAM does Llama 3 8B Instruct abliterated need?
About 10.1 GB at Q8_0 with an 8K context window. That is the 8.5 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 Llama 3 8B Instruct abliterated on an 8 GB GPU?
Yes, with Q4_K: the 4.9 GB file needs about 6.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Llama 3 8B Instruct abliterated GGUF?
It depends on the quant: Q4_K is 4.9 GB, Q8_0 is 8.5 GB, file is 16.1 GB.
What context length does Llama 3 8B Instruct abliterated support?
The GGUF header lists a trained context length of 8,192 tokens (8K). The KV cache grows with the context you use: about 1.1 GB at 8K and 1.1 GB at 8K.
Where can I download Llama 3 8B Instruct abliterated GGUF?
The files are on Hugging Face in the failspy/Llama-3-8B-Instruct-abliterated-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
Can I run Q4_K on 8 GB VRAM?
Yes, Q4_K requires about 6.5 GB memory at 8k context, fitting within 8 GB VRAM.
Does this model support vision tasks?
No, it is a text‑only model with no vision capabilities.
What is the context length?
The model supports up to 8192 tokens of context.