Switch models mid-conversation
The model selector flyout lets you switch the model driving the current conversation at any point, including mid-thread. The change takes effect on the next message you send.
Open the model selector
Section titled “Open the model selector”In any conversation, find the model name displayed in the conversation toolbar or header area. Click it to open the flyout.
The flyout has:
- A title bar showing Model and a count of how many models are currently enabled.
- A subtitle: “Pick a model for this chat. Only what you have connected shows.”
- A search box.
- One or more zones of model rows.
- A More models section (collapsed by default) with additional available models.
- A Manage models link at the bottom.
What the flyout shows
Section titled “What the flyout shows”The flyout organizes models into zones by provider and relevance:
- Flux Auto hero (if Flux Router is connected): a highlighted row at the top for the recommended auto-routing model, labeled “Recommended”. Select it to let Flux Router choose the best model for each turn.
- Pinned models: models you have pinned appear in a dedicated zone above the full provider list.
- Recently used: a short list of the models you have used most recently, capped to four entries.
- Provider zones: all remaining models grouped by provider, marquee providers listed first.
- More models: additional models collapsed under a toggle.
Each row shows the model name, a provider tag, and a price tier badge ($, $$, or $$$) where known. The currently active model has a checkmark.
If no provider is connected, the flyout shows an empty state with a Connect a provider button.
Switch the model
Section titled “Switch the model”- Open the flyout.
- Find the model you want. Use the search box to filter by model name or provider name.
- Click the row. The flyout closes and the conversation header updates to the new model name.
- Send your next message. The agent uses the new model from that point on.
Search for a model
Section titled “Search for a model”Type in the search box to filter the full list. Search matches on model name, provider name, and the model’s descriptor. Results are grouped by provider.
Pin a model
Section titled “Pin a model”Click the pin icon on the right side of any model row to pin it. Pinned models appear in the Pinned zone at the top of the flyout for faster access.
Click the pin icon again on a pinned model to unpin it.
Effort level (where supported)
Section titled “Effort level (where supported)”If the active backend supports an effort or reasoning level (Wayland Core, Claude Code, and Codex), a sub-row appears below the active model row. Use it to choose between lighter and heavier reasoning passes. The setting persists per conversation.
Response length (output budget)
Section titled “Response length (output budget)”Alongside the model and its effort level, you can set how long a response is allowed to be. Choose Auto to let the engine size the right token budget for whatever model is active, or set a fixed budget when you want shorter, capped replies. Auto is the default and the right choice for most work: the engine picks a sensible per-model output budget for you, so there is no model-name guessing and no manual max_tokens to keep in sync as you switch models. The setting follows the conversation, so a fixed budget you set stays in effect until you change it back to Auto.
Picker ordering and filtering
Section titled “Picker ordering and filtering”The picker lists the newest models first within each provider, and it only shows models you can actually chat with: embedding, image, and audio models are filtered out of the chat and workflow pickers. If a provider serves a subscription catalog (such as a ChatGPT subscription), the picker pulls that catalog live from your connected account.
Use a custom model ID
Section titled “Use a custom model ID”If a provider serves a model the catalog has not caught up to yet, you can add a custom model ID for that provider from the providers settings page. The model then appears in the picker like any other row.
When a conversation hits the context ceiling
Section titled “When a conversation hits the context ceiling”If a chat stops because the active model’s context window is full, Wayland offers a one-click model switch right there: pick a larger-window model from the offer and continue the same conversation. The context meter is sized from the model’s real window, and it resizes when you switch models.
Manage models and providers
Section titled “Manage models and providers”Click Manage models at the bottom of the flyout to navigate to the providers settings page. From there you can connect new providers, disconnect existing ones, and configure model availability.
See Add a Cloud Model Provider for the provider connection flow.