Models & the Router
CrewLabs Desktop gives you a curated list of models and a router that picks among them for you. You set this in the setup wizard and can change it any time.
Auto — the smart default
Auto is the default, and it's the right choice for most people. Instead of locking you to one model, Auto reads each message and picks the model that fits it: a quick model for a simple reply, a powerful one for a hard problem. You get good answers on the hard messages without paying for heavy models on the easy ones.
You don't have to manage any of this. Send your message and Auto decides. If you want to see which models it reached for, the per-model usage breakdown on the Credits screen shows the split.
Keep Auto unless you have a specific reason to pin a model.
Pinning a model
If you'd rather the agent always use one model — for consistency, or because you prefer a particular one's style — pin it. Once pinned, every message goes to that model until you change it.
The model list
| Model | Good for |
|---|---|
| Auto | The default. Picks the right model per message. Start here. |
| Claude Opus | The hardest problems — deep reasoning, long multi-step work, careful writing. |
| Claude Sonnet | A strong all-rounder. Capable on most work without reaching for the top tier. |
| Claude Haiku | Fast, lightweight replies. Quick questions and simple tasks. |
| Qwen 3.7 Max | A capable general-purpose alternative. |
| GPT-5.5 | A strong general-purpose model. |
| GPT-5.5 Pro | The heavier GPT tier for demanding tasks. |
"Latest" labels in the app track the newest version of a model, so picking one keeps you current as new versions ship.
Which should I pick?
- Not sure? Leave it on Auto.
- Want maximum capability on everything, cost aside? Pin Claude Opus or GPT-5.5 Pro.
- Mostly quick, simple exchanges? Pin Claude Haiku for speed.
- Want one consistent voice across every reply? Pin whichever model you like best.