Models
Models and model selection
Choose among the model options configured for your Pixel deployment.
What a model is
A model is the AI system that processes a request and generates a response. Models differ in speed, reasoning behavior, supported input types, and provider limits.
Auto selection
Auto uses Pixel’s backend routing policy rather than pinning requests to a single model. The current frontend describes a text route that prioritizes GPT OSS 20B and an image route to a vision model, with fallbacks when available. Routing can change as the backend configuration changes.
Models in the current registry
The built-in frontend registry currently includes Gemini 3.5 Flash, Llama 3.2 11B Vision, Qwen 3.8 27B, GPT OSS 20B, Gemma 4 26B, and Nex N2.5 Mini. The backend determines which options are actually available at runtime.
Availability
If a model is missing or fails, use Auto or another available model. Provider health, credentials, and deployment configuration affect availability.