Models
Add the models your organization offers members, each billed to one of your provider keys.
How-to guide Updated 19 Nov 2026Who can do this
Organizations on Enterprise Cloud. Owners and admins.
What you get
A registered model is one your members use in the PandaOS app. It has a name you choose and is billed to one of your provider keys. You register models on Models in the console. A new model is offered to every team unless you turn that off, see Team model access.
Models has three tabs: Chat, Voice and Image. These steps use Chat. Voice and image models are registered on their own tabs, Voice and Image. They are registered one at a time and need an OpenAI or Azure OpenAI key. For a voice model the dialog adds What it does, and a speech out model adds Voice.
Register a model
You need: an active provider key, see Provider keys. Each model is billed to a provider key. If you have none yet, Register model opens on Add a provider first. Add the key there, then go straight on to choosing the model.
- Open Models in the console and click Register model.
- Choose a Provider key.
PandaOS asks the provider which models this key can reach and lists them. For Azure, the list shows the deployments in your Azure resource. For Google Vertex AI, it shows Model Garden. To add a key without leaving the dialog, choose Add a provider key… in this menu. - Click a model in the list.
If the model is missing, click Enter manually and type the model ID the provider expects. For Azure, type the Deployment name. - Enter the Name members see.
This is how the model is listed in the PandaOS app. Up to 80 characters, and no two models can share a name. - Choose Where it runs: Global, EU or Germany.
Germany also counts as EU. Members who filter by hosting do not see a model left on Not set. For Google Vertex AI, the key's location fills this in when PandaOS knows the region. The Runs in column of the models list shows it for every model. - Leave Available to every team on, or turn it off.
On, every team gets the model, including teams created later. Off, you switch it on per team in Team access. - Click Register model.
Register several models at once
- On Models, click Register model and choose a Provider key.
- Tick each model you want.
Up to 25 models in one registration. - Click Review with the number of models you ticked.
- Check the Name members see for each model, and enter both prices for any model marked "Unpriced".
- Click Register with the number of models.
Each model is saved on its own and offered to every team. If some are refused, the dialog stays open on those rows with the reason. Fix them and register again.
What you see when it worked
The model is listed on Models, under Chat models. The console shows a status only while something is under way or needs you. The Sync column reads Queued or Syncing while the model is on its way, and shows nothing once it is live. The page refreshes every 10 seconds while a change is in progress.
The Teams column shows the All teams switch, and a note only when something is wrong for a team, for example "Syncing for 1 of 3 teams" or "Not offered to any team". Members with access get the model within a few minutes, under the name you chose.
Register model cannot be clicked
Register model is greyed out only when the model list did not load. Reload the page.
Without an active provider key the button still works. It opens Add a provider first, where you add a key before you pick the model. See Provider keys.
The model list does not load
Click Try again. If the list still fails, click Enter manually and type the model ID the provider expects.
A model with this name already exists
Two models cannot share a name. Choose another Name members see.
The Sync column says Failed
Hover over Failed to see the reason and when PandaOS tries again. It retries on its own. If the reason points at the model, such as a wrong deployment name, click Edit in the model's menu and correct it.
Sync status
| Status | Meaning |
|---|---|
| Queued | Waiting to be sent. |
| Syncing | Being sent. |
| No status | Live for members with access. |
| Failed | The last attempt failed. Hover to see the reason and the time of the next retry. |
| Removing | Being taken off. |
| Removed | Taken off. |
Prices and advanced settings
Open Advanced in the dialog for these settings.
| Setting | What it does |
|---|---|
| Input price and Output price | USD per million tokens. Only needed when PandaOS cannot price the model. When it has no price at all, both prices are required. |
| Max input tokens | The longest prompt the model takes. |
| API version | Azure only. Leave empty to use the default. |
| Location | Google Vertex AI only. The region the model runs in. |
| Accepts images | Members can attach pictures to prompts. |
| Custom regions | Your own hosting tags, for example a cloud region. A tag you add once is offered on every model. |
Enable, disable or remove a model
Use the Enabled switch on the model's row, or its menu.
| Action | What happens |
|---|---|
| Disable | Members cannot use the model until you enable it again. Its team access is kept. |
| Enable | The model comes back for every team that still has access. |
| Remove | The model is deleted, together with its team access. Registering it again starts from scratch. |
| Offer to all teams | Every team gets the model, including teams created later. Stop offering to all teams takes it from no one. |
If a model tier is pinned to a model you disable or remove, the console warns you first. That tier falls back to PandaOS picking a model. See Model tiers.
To change a model's name, hosting, prices or provider key, click Edit in its menu. A disabled or removed model disappears from members' model picker once the change is applied.