What's the Point of Local AI?
10%
A Sensible Default
Why not use both, open and closed models?
If a multi-GPU rig for the big open models is out of reach, run a small open model on your own machine for the bulk of your work and keep a paid API key for the hard tasks where the best model earns its price.
Locally, start with a model that fits a normal GPU: something like gpt-oss:20b
Local AI Engineering with Ollama
Run, understand, customize, fine-tune, and build agentic apps on your own hardwareEnroll now to unlock all content and receive all future updates for free.