Choosing models
Harriet Desktop gives you every leading model your organization has enabled, in one picker, with no personal AI subscription. This page covers how to pick a model per chat, where the list comes from, and what to do when it's empty.
The model picker
The picker works at two levels:
- Chat model — choose the model for the current chat. If a model supports reasoning profiles, you configure them on its card.
- Default model — choose the model used for new chats, set from Settings.
The picker groups Recommended models first and is searchable, so you can jump straight to the one you want. Match the model to the job: a heavyweight reasoning model for a finance forecast or an IT architecture review, a fast model for a sales rep churning through call summaries or an operations team reformatting shift notes.
Where the list comes from
Model traffic from Harriet Desktop goes through Harriet's LLM proxy rather than to each provider directly. Your organization supplies the credentials behind it — its own provider keys (BYOK) or Harriet's managed credits — and Harriet records usage and cost.
Which models you see is a policy decision, not a personal one: the picker shows exactly what your organization's model allowlist permits. Everyone signs in with their Harriet account and gets the same governed list, so there is no per-seat AI subscription to buy or manage.
Because the proxy carries the credentials, you never paste an API key into Harriet Desktop, and switching providers never requires new accounts on your machine.
If the picker is empty or missing a model
When no models have reached the app yet, the picker says so directly: "No LLM models are available from your Harriet account yet. Sync from Settings → Harriet account, or check LLM billing and model access in the Harriet console."
- Sync first. Open Settings → Harriet account and sync; a fresh sign-in or a newly enabled catalog can take one sync cycle to land.
- Check the allowlist. A model that exists but isn't listed for you is usually excluded by the allowlist — that's an admin setting, not an app fault.
- Check billing. If your organization's keys or credits aren't set up for a provider, its models can't be served. Model availability walks through the cases.