Local models
A local model is a brain that runs on your own computer: free forever, private by physics, working even when the internet is not. This guide is the Local Models page, start to finish.
narrated · sound on
Why this matters
Per-token bills creep. A free local floor means the expensive brains only wake for work that earns them - and the private stuff never has to leave home at all.
The walkthrough
Install the engine once
Local models run through Ollama. Install it once and the Local Models page lights up on its own - Cronus detects it and takes over the management.
Pull a model
Type a model name (like qwen3:4b - the tag is the size, four billion parameters) and hit Download. Progress shows on the page; a few minutes later the model is yours, permanently, offline.
Read the fit badges
Every installed model gets an honest badge against your machine's actual memory: fits, will be slow, or will not fit. Measured, not guessed - so you know before you commit.
Use it everywhere
Pulled models appear in the model picker marked local, free. Rule of thumb: local for everyday questions and routing, flat-rate cloud for the daily grind, frontier models only for the hardest work. That routing discipline is where the savings live - and in Cronus it is a setting, not a habit you have to maintain.