AI Models
AppX is powered by Google's Gemini 3.1 family. You don't pick a model or tune anything — AppX picks the right one for each step automatically, so you can stay focused on describing your app.
What's under the hood
AppX leans on two strengths of the Gemini 3.1 family:
- Gemini 3.1 Pro — the architect and builder. When AppX plans your app and writes the React Native code for each screen, it uses the most capable model so the structure holds together and the code follows real Expo + React Native patterns.
- Gemini 3.1 Flash — the fast hand for edits. When you refine an app through chat, AppX uses the faster model to apply precise, surgical changes and quick fixes. It's quick and light, which keeps your edit loop snappy.
You get the strong model where quality matters most and the fast model where speed matters most — without ever flipping a switch.
Why you don't choose
Picking models is a chore, and the right choice changes from one action to the next. AppX makes that call for you on every operation, so a fresh app build, a new screen, and a one-line tweak each run on the model best suited to the job.
Big context, kept lean
These models can hold a lot of context at once, but AppX deliberately sends only what's relevant — the screen you're working on, plus short summaries of the rest. That keeps every request fast and your credits efficient, which is part of why edits feel quick.
Always current
The AI landscape moves fast, and so does AppX. When a newer model is clearly better — faster, sharper, or more efficient — AppX moves to it so your apps get better without any work on your side.