Run Smaug Llama 3 70B Instruct Abliterated v3 on Mac
Uncensored abliterated Llama 3 70B retaining original knowledge
Smaug Llama 3 70B Instruct Abliterated v3 runs 100% private on Mac inside Private LLM — no internet connection required, no data sent to any server.
Model on Hugging Face
Specifications
- Parameters
- 70.6B
- Context window
- 8K tokens
- License
- Llama 3
- Quantization
- OmniQuant 3-bit, OmniQuant 4-bit
- Family
- Llama 3 70B
What Smaug Llama 3 70B Instruct Abliterated v3 is good at
A 70.6 billion parameter model, Smaug Llama 3 70B Instruct Abliterated v3 removes the refusal direction from the original Smaug Llama 3 70B Instruct weights via orthogonalization. This leaves other behaviors and knowledge unchanged while making the model far less likely to reject sensitive or unrestricted prompts. It is not fine-tuned, belongs to the Llama 3 70B family, and handles up to 8192 tokens of context. Occasional refusal or safety warnings may still appear.
Which of your devices can run it
Mac
How to run Smaug Llama 3 70B Instruct Abliterated v3 in Private LLM
- Download Private LLM from the App Store.
- Open the in-app model library and choose Smaug Llama 3 70B Instruct Abliterated v3.
- Download the model once, then chat fully offline.
Variants & related models
Cat Llama 3 70B Instruct
Immersive roleplay via strict prompt adherence, Llama 3 70B finetune
Smaug Llama 3 70B Instruct
General conversationalist fine-tuned from Llama 3 70B, matching GPT-4 Turbo dialogue
Amoral Gemma 3 1B v2
Uncensored Gemma 3 1B finetune for neutral factual responses
DeepSeek R1 Distill Llama 8B Abliterated
Uncensored DeepSeek R1 Distill 8B via refusal abliteration
DeepSeek R1 Distill Qwen 32B Abliterated
Uncensored proof-of-concept removing alignment guardrails from DeepSeek R1 Distill
Frequently asked questions
Smaug Llama 3 70B Instruct Abliterated v3 is too large for current iPhones. It runs on Mac with enough unified memory inside Private LLM, fully offline.
Yes. Smaug Llama 3 70B Instruct Abliterated v3 runs on Macs with enough unified memory, such as Mac (Apple Silicon, 192GB), MacBook Pro (M4 Max, 128GB), Mac Studio / Pro (Apple Silicon, 96GB), MacBook Pro (M4 Max, 64GB), MacBook Pro (M4 Max, 48GB), fully on-device in Private LLM.
Yes. Once downloaded in Private LLM, Smaug Llama 3 70B Instruct Abliterated v3 runs 100% on-device — no internet connection, and nothing is sent to any server.
Private LLM is a one-time purchase with no subscription and no per-message cost. The models themselves are open source — once downloaded, they run offline with nothing to pay per use.
Why run Smaug Llama 3 70B Instruct Abliterated v3 in Private LLM
Private LLM has run local AI on iPhone, iPad, and Mac since 2023, well before Apple Intelligence, LM Studio, Ollama, etc. existed. Inference happens on your device, so your Smaug Llama 3 70B Instruct Abliterated v3 conversations never reach a server. The part most apps gloss over is quantization, and that is exactly where on-device quality is won or lost. Most llama.cpp and MLX wrappers ship the same off-the-shelf 4-bit RTN weights. Private LLM ships GPTQ and OmniQuant quantization, tuned per model, and our 3-bit OmniQuant models match or beat those 4-bit RTN builds on the same Apple Silicon. Run the same model both ways and you feel it in the first reply. See how our quantization works.
Specifications and summary come from Smaug Llama 3 70B Instruct Abliterated v3's Hugging Face model card, released under the Llama 3 license. Private LLM ships its own quantized models, built with OmniQuant quantization tuned per model, and isn't affiliated with the model's authors.