The model selector is in the message box, on the name of the model in use. You use it to choose the model that answers, to let “Auto” choose for each question, to turn on agent mode and to compare several models on the same request.
Open the selectorLink to the Open the selector section
- Click the model name, in the message box.
- Or press CtrlShift2 (Windows, Linux) or ⌘Shift2 (Mac), on the website as well as in Moon Desktop. The shortcut is shown in the selector’s search field.
On a computer, the selector opens as a menu above or below the message box, on the side with the most room. The list always shows at least six to seven models: when neither side has enough room, for example in a short window with the message box in the middle of the home screen, the selector opens as a centered window, with the same views and the same shortcuts.
On a phone or a touch tablet, it opens as a sheet from the bottom of the screen, almost full screen. A single row of filters scrolls sideways: “Favorites”, “All”, then the filters by type. The sheet opens on “All” when you have fewer than six favorites. While the keyboard is open on the search, the bottom of the selector steps aside to leave the room to the list.
What’s in the selectorLink to the What’s in the selector section
From top to bottom:
- The “Search … models” field.
- A row of filters: “All” (on a phone, “Favorites” and “All”), then, depending on the models offered, “Fast”, “Reasoning”, “Code”, “Vision” and “My models”.
- The “Auto” row: “Moon AI picks the best model for each question” (see Auto mode).
- Your favorites: your pinned models or, if you have none, a short selection.
- The models, by section: “Recently used”, “Your models and agents”, “Models of your organization”, “Shared with you”, “Moon AI”, one section per provider, then “Other”.
- “Previous versions”: the older versions of a model, collapsed.
The bottom of the selector fits on one line, to leave the room to the list:
- On a computer: the “Agent” switch, “Approve the plan before starting” in agent mode (see Agent mode), “Thinking and style” with the current choice, then on the right “Compare two models”, “Set as default” (or the “Default model” check mark) and “Model options” (“…”).
- On a phone: “Agent”, “Thinking and style” with the current choice, “Compare two models” and “Model options” (“…”), which holds “Set as default”. In agent mode, “Approve the plan before starting” has its own line above.
For a model whose thinking can’t be adjusted, “Thinking and style” becomes “Style” (see Answers).
The markers on each modelLink to the The markers on each model section
| Marker | Meaning |
|---|---|
| “New” | The model came out recently. |
| Three bars: “Low cost”, “Medium cost”, “High cost” | The model’s relative cost. The higher it is, the faster your usage limit fills up. |
| “Very high cost” | A single message uses a large share of your 5-hour limit. The percentage shows on hover. |
| “Available with …” or “Not included” | The model isn’t included in the plan of the open workspace. |
Moon AI’s in-house models are in the “Moon AI” section, with Moon’s crescent. The catalog changes often: the current list is in the selector and on the Models page. The Free plan offers a selection of models: see the plans.
Search and filterLink to the Search and filter section
In the search field, type the name of a model, of a provider, or a specialty. Specialties are searched in your interface language: “writing” finds the writing models, “legal” the models specialized in law.
Recognized specialties: “Fast”, “Code”, “Reasoning”, “Vision”, “Writing”, “Documents”, “Multilingual”, “Web search”, “Legal” and “Security”.
Clicking a filter narrows the list, a second click removes it. “Clear search” empties the field.
With the keyboard, the arrow keys move through the list, Enter picks the highlighted model and Esc goes back, then closes the selector.
Previous versionsLink to the Previous versions section
When a provider releases a newer version of a model, the older one moves to “Previous versions”. “Show more, N older models” expands them, “Show less” collapses them. They can still be used like the others.
Pin a modelLink to the Pin a model section
- Choose the model in the selector.
- Open “Model options” (“…”), at the bottom of the selector.
- Click “Pin to sidebar”.
The model appears in your favorites, at the top of the selector, and in the sidebar. “Unpin from sidebar” removes it. The same menu offers “Copy link” to the model and, for your own model or agent, “Edit”.
Set the default modelLink to the Set the default model section
The model you choose applies to the current conversation. “Set as default” keeps it for your new chats: at the bottom of the selector on a computer, in “Model options” (“…”) on a phone. On a computer, when that is already the case, a “Default model” check mark replaces the button.
When the model of a conversation has been removed from the catalog, the message box says so: “The model used in this conversation is no longer available. Pick another one.”
Choose the model for a single messageLink to the Choose the model for a single message section
Type @ in the message box, then the start of a model name. The mentioned model answers this message only; the conversation’s model doesn’t change. Up to three mentioned models answer side by side. See the @ command.
Compare modelsLink to the Compare models section
- Open the selector and click “Compare two models”.
- Choose the second model (“Pick a model to compare”).
- Send your message. The answers appear side by side.
During a comparison, the selector shows one row per model, “Add a model” for one more column and “Stop comparing”. Each answer counts toward your usage limit. Agent mode and “Auto” mode don’t apply during a comparison.
The model councilLink to the The model council section
The council sends your question to three models from three different providers, then writes a synthesis: the best answer, the points of agreement and the differences.
- Open the “+” menu of the message box, then “More” › “Model council”.
- Write your question and send it.
- The three answers arrive, then the “Council synthesis” is written above them. “Show original answers” expands them.
The council uses about four answers’ worth of your limit. If the synthesis couldn’t be written, “Write synthesis” tries again; “Rewrite synthesis” produces a new one. The rest of the conversation continues from the synthesis.
The council is grayed out, with its reason, in agent mode, during a comparison, with a model mentioned with @, or when your plan doesn’t give access to models from three providers.
Very high cost modelsLink to the Very high cost models section
When you choose a model marked “Very high cost”, a notice appears above the message box: “One message uses about …% of your 5-hour limit”, or “One message can use your whole 5-hour limit”.
| Notice button | What it does |
|---|---|
| “Switch to Auto” | Moon AI picks a suitable model for each question. |
| “Let Auto choose” | Shown when you mentioned this model with @ in “Auto” mode: the mention is removed. |
| “Hide for this model” | The notice no longer appears for this model. |
In agent mode, a request often uses several times more. To track your consumption, see Usage limits.