Llama
by Meta
Llama is Meta's family of open-weight models. Because the weights can be downloaded, developers run them on their own hardware, fine-tune them, or use them through many hosting providers. Meta has since introduced other model families too, so check Meta's site for what is current.
- Running AI locally or privately
- Fine-tuning and research
- Avoiding vendor lock-in
Check the vendor's own site for current plans, pricing and limits. Those change often.
Guides
Prompts
Log Line Classifier
A closed label set and a fixed output format suit local models and make results easy to scan or paste into a spreadsheet.
it-ops
Private Document Summary on Your Own Machine
Running a model locally can keep sensitive text on your own computer, though you should confirm your setup does not send data elsewhere.
productivity
Short Commit Message Writer
Short, tightly constrained prompts like this work well with smaller local models that can lose focus in long instructions.
coding
Test a Local Open Model Fairly
Five small tasks that you can judge yourself reveal summary quality, format discipline, coding, explanation and reasoning.
learning
Questions & answers
Agent templates
Phishing Email Analyzer
Helps staff judge a suspicious email by walking through sender, links, urgency and attachments, then says what to do. Defensive use only.
4 min read
Sysadmin Log Analyzer
Reads pasted log excerpts, groups the errors, builds a timeline and ranks likely causes with the next read-only checks to run.
4 min read