Dashboard: Models
The Models page lets you register, edit, clone, and remove LLM models. All models registered here become available for use in router routing configurations.
Health status is shown inline in the same table as configuration -- no separate tab is needed.
Model List

Filtering and Search
- Provider filter -- dropdown to show models for a single provider (OpenAI, Anthropic, Ollama, etc.) or all providers.
- Search -- text search by model ID, filtered live as you type.
Results are paginated at 20 models per page.
Model List Columns
| Column | Description |
|---|---|
| ID | Provider model identifier |
| Provider | Provider badge |
| Endpoint | Base URL used for this model |
| Input $/1M | Input token price in USD |
| Output $/1M | Output token price in USD |
| Cache $/1M | Cache read price (if applicable) |
| Context Size | Maximum context window tokens |
| Status | Healthy, Unavailable -- based on recent error rate. Shows a neutral badge when no traffic has been recorded yet |
| Error Rate (5M) | Percentage of failed requests in the last 5 minutes |
| P95 Latency (5M) | 95th-percentile response time in the last 5 minutes |
| Requests (1H) | Total requests in the last hour |
| Last Success | Time of the most recent successful request |
| Cooldown | Remaining cooldown if the model triggered rate-limiting |
Health columns show a dash when no data is available for the window. The health data refreshes automatically every 30 seconds (visible in the subtitle bar).
With models:write, clicking a row on the Models tab opens that model's edit form, the same page the pencil icon opens; without the permission the row is inert. Rows on the Health tab are not clickable: their only action is the per-row resilience reset button.
Click any column header to sort.
The standalone Provider Health page has been merged into this page. /dashboard/health redirects to /dashboard/models.
Adding a Model
-
Click + Add Model
-
Fill in the form:
- Model ID -- the identifier sent to the provider (e.g.
gpt-5-mini) - Provider -- select from the dropdown
- Connection details -- choose Preconfigured or Custom; see Preconfigured vs Custom Connection below
- Context Window -- pre-filled for known models
- Pricing -- input/output/cache prices per 1M tokens; pre-filled for known models
- Pricing Tiers -- add a tier for long-context pricing (e.g. Anthropic above 200k tokens)
- Capabilities -- check all that apply
- Model ID -- the identifier sent to the provider (e.g.
-
Click Save
Preconfigured vs Custom Connection
The model form's "Connection details" section is a Preconfigured / Custom toggle, not a fixed API Key + Base URL pair:
- Preconfigured -- pick an existing connection from a
dropdown, filtered to connections whose provider matches the model's
Provider. No endpoint or credential fields are shown; none are sent.
The model binds to that connection's
connectionId. If no connection exists yet for the selected provider, the dropdown is replaced with a message pointing you to the Connections page. Picking acustomconnection that carries an upstream provider name fills the model's Provider (upstream provider name) field with it, so the model ID prefix follows the connection. - Custom -- the classic inline Base URL and API Key fields (plus
the provider-specific fields for Azure/Bedrock/Vertex/web providers), typed
directly into the model form. On save, Routerly creates or updates a
dedicated, single-model connection with id
conn-for-<model-id>from those values and binds the model to it. This is functionally identical to the pre-cutover behavior; the credentials still end up on a connection under the hood.
Both modes produce a model that resolves through a real connection -- Preconfigured just lets more than one model reuse the same one.
Editing a Preconfigured connection on the Connections page updates the endpoint and credentials for every model bound to it, immediately. There is no per-model override once a model is on a shared connection. If a model needs its own credentials, switch it to Custom instead.
Editing an existing model defaults the toggle based on how the model is currently bound:
- Bound to a connection shared with other models (any
connectionIdother than that model's ownconn-for-<model-id>) -- defaults to Preconfigured, with that connection pre-selected. - Bound to its own dedicated
conn-for-<model-id>connection -- defaults to Custom, with the Base URL pre-filled and API Key left blank (never echoed back).
Use the Model Discovery page (navigate to Models, then click Discover) to browse the built-in catalog. Clicking Add next to any entry opens the new-model form with the Provider and Model ID already filled in, and pre-populates pricing and context window from the catalog entry. You only need to supply the API key.

Editing a Model
Click the Edit (pencil) icon next to a model. All fields except the Model ID are editable.
The Connection details toggle opens on Preconfigured or Custom
depending on the model's current binding -- see
Preconfigured vs Custom Connection. In
Custom mode, to update the API key enter a new value -- Routerly re-encrypts
it immediately and writes it to the model's dedicated conn-for-<model-id>
connection. Leaving the API key field blank keeps the currently stored
credentials unchanged.
Catalog Tracking and Field Overrides
Models linked to the catalog show status badges on editable fields:
- Auto (green badge) -- field is synced from the catalog; your edit will override it
- Override (amber badge) -- field is manually locked to your custom value; it will not auto-sync
- No badge -- model is not in the catalog (local Ollama, custom endpoint) or field is always manual (API key, enabled status)
To override a catalog field: edit it normally. The badge changes to "Override" and the field is locked. An edit indicator appears showing the catalog default value.
To reset an overridden field: click the Reset button next to it. The field reverts to the catalog value and the override lock is removed. Auto-sync resumes.
The override status is saved with the model; if you edit the same model again, your locks persist.
Cloning a Model
Click the Clone icon to create a copy of a model entry. Useful when registering a fine-tuned variant that shares the same provider and pricing as a base model.
Change the Model ID and API Key as needed, then save.
Disabling a Model
Toggle the Enabled switch to off to temporarily remove a model from routing without deleting it. Disabled models are visible in the list but are excluded from all routing decisions.
Removing a Model
Click the Delete (trash) icon. You will be asked to confirm.
Removing a model that is assigned to active router routing configurations will cause routing failures for those routers. Remove the model from all router routing configs before deleting it.
Model Discovery
Navigate to /dashboard/models/discover (or click Discover on the Models page) to browse the built-in model catalog. The catalog lists known models from supported providers with their context window and published pricing.

| Column | Description |
|---|---|
| Model | Model ID (sortable) |
| Provider | Provider name (sortable) |
| Context | Maximum context window, e.g. 128k, 1.0M (sortable) |
| Input /1M | Input price per 1,000,000 tokens; free for zero-priced models (sortable) |
| Output /1M | Output price per 1,000,000 tokens (sortable) |
Click any column header to sort ascending; click again to reverse. All five columns are sortable.
Filters (two rows):
- Search -- filter by model ID or name substring
- Provider -- multi-select dropdown; leave empty to show all providers
- Context -- limit by context window size: All, <32k, 32k–200k, 200k–1M, >1M
- Price /1M -- filter by input price tier: All, Free, <$1, $1–$5, >$5
- Show -- All, Configured only (models already added to Routerly), Embedding only
Click Reset filters to clear all filters at once.
Click Add next to any model, or anywhere else on its row, to open the new-model form with that model's details pre-filled. Pricing and context window are populated automatically from the catalog. You only need to supply the API key to complete registration.