Getting started
Model and effort per app
Pick the model and effort for each window or conversation in every Jennefer app, set a default for each app in Hub > Models > Each app, and see which session uses which model.
On this page
Every Jennefer app that uses a model lets you pick the model and the effort for each window or conversation. Two IDE windows or two CLI terminals can use different models at the same time.
Which model a request uses#
Three levels decide. The first one that is set wins.
| Level | Where you set it | Used by |
|---|---|---|
| Session choice | The app's own picker, for example the IDE model chip or /model in the CLI. | That window, conversation or chat only. |
| App default | Hub > Models > Each app, or Make IDE default, /model save and Make this the … default in the apps. | New windows and conversations of that app. |
| Hub default | The top card in Hub > Models. See Models. | Every app without an app default. |
Every picker lists the models on this computer and, once your endpoint is approved, your saved endpoint models in blue. See Your endpoint.
Effort#
Effort sets how long the model thinks before it answers. Default keeps the model's own setting.
| Effort | Model on this computer | Endpoint model |
|---|---|---|
| Off | Thinking is off. | Not used. |
| Low, Medium, High | Thinking is on. | Sent to your provider as the effort level. |
| Extra high, Max | Thinking is on, the same as High. | Used by Claude on Anthropic and Amazon Bedrock. |
Higher effort is slower and uses more output tokens. On your endpoint, your provider bills those tokens.
Hub > Models > Each app#
Each app lists IDE, CLI, CLI+, Workshop, Hangar, Cloud and Mobile. Each row has a model picker and an effort picker.
- Hub default keeps the app on the Hub default. Pick a model to give the app its own default.
- The effort picker works once the app has its own model.
- If you delete the model an app uses, its row says The saved model is gone; Hub default is used.
- Connected apps are not in this list. They use the model on this computer, never your endpoint.
Open now#
Open now lists every window, conversation and bot that asked a model something in the last two hours: the app, the agent or session, the model and effort, own pick or default, and when it was last seen. When nothing ran, it says No app has asked a model anything in the last two hours.
Pick in each app#
| App | Where you pick | Save as the app default |
|---|---|---|
| IDE | The model chip in the header. This window lists the default, the models and effort chips. | Make IDE default |
| CLI | /model opens a picker. /model <name> picks directly, /model default goes back to the default, /model hub opens Hub > Models. /effort <level> takes off, low, medium, high, xhigh, max or default. The choice is saved with the conversation. | /model save |
| CLI+ | /model lists the models. /model <number or name> and /model default change it, /effort [level] sets the effort. The HUD shows the effort. | /model save |
| Workshop | The model chip next to the box where you describe a new app or a change. | Make this the Workshop default |
| Cloud | The model chip next to the team's message box. | Make this the Cloud default |
| Hangar | Jen's model chip under her message box. Each agent step of a bot has a Model field; left on the default, the step uses the Hangar default, then the Hub default. | Make this the Hangar default in Jen's chip |
| Mobile | The model chip in a chat's header opens Model for this chat: Computer default, the models on your computer, your endpoint models, models In your account, and Effort. A new chat starts with the current chat's model. | Hub > Models > Each app > Mobile |
Context window and thinking for models on this computer are set in Hub > Models. Each local model keeps its own settings.
Models on this computer and endpoint models#
- Local models share one GPU and take turns. Requests for the same model share it. A request for another model waits until the running answers end, then the Engine switches models. See GPU.
- While a request waits, the app says so, for example Waiting for the GPU: CLI is answering with Gemma 4 12B or Loading Qwen3.5 9B (about 20 s).
- Endpoint models do not use the GPU queue. Their requests run in parallel, also next to local models.
FAQ#
Does picking a model in one window change my other windows?
No. A pick applies to that window, conversation or chat only. Make … default changes what new windows start with.
Why is my endpoint model not in the picker?
Endpoint models show only after your endpoint is approved and a model is saved in Hub > Models > Your endpoint. See Your endpoint.
Where do I set the context window?
For local models in Hub > Models. For endpoint models, choose Edit on the saved model in Your endpoint.
Can I see what each app costs?
Yes. Endpoint models show an estimated cost in the apps and in Hub > Usage.