Meta-Llama-3-70B-Instruct-abliterated-v3.5 needs about 35.1 GB of memory at IQ3_M (a 31.9 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 25 quants, from 15.3 GB (IQ1_S) to 141.1 GB.
Meta‑Llama‑3‑70B‑Instruct‑abliterated‑v3.5 – a 70B instruction model with orthogonalized refusal weights
This 70‑billion‑parameter Llama model has been orthogonalized to reduce refusal behavior while keeping all other instruction capabilities intact. It retains the original 8192‑token context window and is distributed in multiple GGUF quantizations. The model is licensed under the llama3 license.
Good at
- Large‑scale instruction following
- Reduced refusal signals
- High‑quality text generation
Limits
- Requires large VRAM for high‑quality quants
- No vision or multimodal support
- License restricts commercial use
Best for
- Chatbot back‑ends needing minimal refusal
- Research on refusal mitigation
- Fine‑tuning for custom instruction sets
Who it’s for. Developers and researchers who need a 70B instruction model with reduced refusal behavior. The model is suitable for local deployment using llama.cpp and for experiments on orthogonalization techniques.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_S | 15.3 GB | 18.5 GB | Too big | Get file → | |
| IQ1_M | 16.8 GB | 19.9 GB | Too big | Get file → | |
| IQ2_XXS | 19.1 GB | 22.3 GB | Too big | Get file → | |
| IQ2_XS | 21.1 GB | 24.3 GB | Too big | Get file → | |
| IQ2_S | 22.2 GB | 25.4 GB | Too big | Get file → | |
| IQ2_M | 24.1 GB | 27.3 GB | Too big | Get file → | |
| Q2_K_S | 24.5 GB | 27.7 GB | Too big | Get file → | |
| Q2_K | 26.4 GB | 29.6 GB | Too big | Get file → | |
| IQ3_XXS | 27.5 GB | 30.7 GB | Too big | Get file → | |
| IQ3_XS | 29.3 GB | 32.5 GB | Too big | Get file → | |
| IQ3_S | 30.9 GB | 34.1 GB | Too big | Get file → | |
| Q3_K_S | 30.9 GB | 34.1 GB | Too big | Get file → | |
| IQ3_M | 31.9 GB | 35.1 GB | Too big | Get file → | |
| Q3_K | 34.3 GB | 37.5 GB | Too big | Get file → | |
| Q3_K_L | 37.1 GB | 40.3 GB | Too big | Get file → | |
| IQ4_XS | 37.9 GB | 41.1 GB | Too big | Get file → | |
| IQ4_NL | 40.1 GB | 43.2 GB | Too big | Get file → | |
| Q4_K_S | 40.3 GB | 43.5 GB | Too big | Get file → | |
| Q4_K | 42.5 GB | 45.7 GB | Too big | Get file → | |
| Q5_K_S | 48.7 GB (2 parts) | 51.8 GB | Too big | Get file → | |
| Q5_K | 49.9 GB (2 parts) | 53.1 GB | Too big | Get file → | |
| Q6_K | 57.9 GB (3 parts) | 61.1 GB | Too big | Get file → | |
| Q8_0 | 75.0 GB (3 parts) | 78.2 GB | Too big | Get file → | |
| BF16 | 141.1 GB (6 parts) | 144.3 GB | Too big | Get file → | |
| — | 141.1 GB (6 parts) | 144.3 GB | Too big | Get file → |
For 8 GB VRAM, use IQ1_S (15.34 GB file, ~18.5 GB memory at 8k context). With 16 GB VRAM, IQ2_XS (21.14 GB file, ~24.3 GB memory) or Q2_K_S (24.47 GB file, ~27.7 GB memory) are suitable. For 24 GB VRAM, Q3_K (34.27 GB file, ~37.5 GB memory) or Q4_K (42.52 GB file, ~45.7 GB memory) provide higher quality.
What GPU do you need for Meta-Llama-3-70B-Instruct-abliterated-v3.5?
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 IQ2_XXS (19.1 GB file, about 22.3 GB in use).
- 32 GB VRAM: the best quant that fits is IQ3_XXS (27.5 GB file, about 30.7 GB in use).
- 48 GB VRAM: the best quant that fits is Q4_K (42.5 GB file, about 45.7 GB in use).
Which Meta-Llama-3-70B-Instruct-abliterated-v3.5 quant should you choose?
IQ3_M (31.9 GB) is the middle option in this repository. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.
If memory is tight, the smallest file is IQ1_S at 15.3 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Frequently asked questions about Meta-Llama-3-70B-Instruct-abliterated-v3.5
How much VRAM does Meta-Llama-3-70B-Instruct-abliterated-v3.5 need?
About 35.1 GB at IQ3_M with an 8K context window. That is the 31.9 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 Meta-Llama-3-70B-Instruct-abliterated-v3.5 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 18.5 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Meta-Llama-3-70B-Instruct-abliterated-v3.5 GGUF?
It depends on the quant: IQ1_S is 15.3 GB, IQ1_M is 16.8 GB, IQ2_XXS is 19.1 GB, IQ2_XS is 21.1 GB, IQ2_S is 22.2 GB, IQ2_M is 24.1 GB and 19 more.
What context length does Meta-Llama-3-70B-Instruct-abliterated-v3.5 support?
The GGUF header lists a trained context length of 8,192 tokens (8K). The KV cache grows with the context you use: about 2.7 GB at 8K and 2.7 GB at 8K.
Where can I download Meta-Llama-3-70B-Instruct-abliterated-v3.5 GGUF?
The files are on Hugging Face in the legraphista/Meta-Llama-3-70B-Instruct-abliterated-v3.5-IMat-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 Meta-Llama-3-70B-Instruct-abliterated-v3.5 use?
Hugging Face lists the license as llama3. 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, the IQ1_S quant requires about 18.5 GB of VRAM at 8k context, so 8 GB is insufficient. Use a lower‑rank quant or a smaller context.
Does this model support vision or multimodal inputs?
No, the model only handles text. It has no vision or multimodal capabilities.
Is the model free to use for commercial projects?
The llama3 license restricts commercial use. Check the license before deployment.