The model picker sits in the message box. Click it to see which AI models your workspace can use. Auto is the default and picks a sensible model for you; you only need this menu when you want to trade speed against depth yourself.
What's in the list
Which models appear depends on your plan and, for self-hosted installs, on what the operator has configured. On Relaticle Cloud:
| Model | Availability | Credit multiplier |
|---|---|---|
| Auto | Everyone | Depends on the model chosen |
| Claude Sonnet | Everyone | 1x |
| Claude Opus | Pro | 3x |
| GPT-5 class models | Pro | 1.5x |
Models marked with an amber Pro badge need a Pro workspace. The multiplier scales what a message costs in credits. See AI credits and rate limits.
Self-hosted workspaces can also point the assistant at their own models (Ollama or any OpenAI-compatible endpoint); those appear in the same picker once configured.
How the choice behaves
Your selection is remembered in the browser you made it in and applies to the messages you send from then on. It's a per-person choice, so teammates pick their own. Switching back to Auto at any time is always safe.