Reference · Glossary

Local LLM

Last updated

A **local LLM** runs on your machine or private hardware instead of a vendor chat API. Useful for privacy, offline experiments, and learning serving tradeoffs.

#When to use

Sensitive prototypes, air-gapped exploration, cost control for high-volume low-stakes tasks, and learning model limits hands-on.

#When not to

Expecting frontier quality on a laptop GPU for hard reasoning, or skipping evals because “it’s private so it’s fine.”

#Quality checklist

  • Hardware fits model size (RAM/VRAM)
  • Pin model revision
  • Compare against a cloud baseline on golden tasks
  • Know quantization’s quality cost

#Learn next