0
Prompts
Meta Runs locally
Llama
Llama runs on your own hardware through Ollama, llama.cpp or LM Studio, which makes it the practical choice for private data and offline work. Smaller quantised variants follow multi-part instructions less reliably than hosted models, so split long prompts into single tasks and keep the output format simple. Parameters live in a Modelfile rather than in the prompt.
1
Configs
Modelfile
Config file
Yes
Runs locally
Models
Llama 3.2 8B
Llama 3.3 70B
Configuration files
All 1- LlamaMistral
Ollama Modelfile for a local code assistant
A Modelfile with a system prompt and parameters tuned for short, correct code answers on a local model.
Guides
- LlamaMistral
Running a model on your own machine
What local models are genuinely good for, what hardware you need, and what to change in prompts that came from a hosted model.
Common questions
Where do you put standing rules for Llama?
In Modelfile. It is read on every session, so anything you would otherwise paste into each message belongs there rather than in the conversation.
Do Llama prompts work on other models?
Usually with changes. Every resource here records where it was tested and what to adjust elsewhere, and portable ones are filed under Universal.
Can Llama run on my own machine?
Yes. Memory is the constraint: a 4-bit quantised model needs roughly 0.6 GB per billion parameters plus room for the context window.