Catalog
The models this caller may use, one entry each however many providers serve it.
Prices are the caller's: an organization's override where one applies, else the deployment's row, else the models.dev default. Aliases and routing policies are not models and are not listed; see Routing. A visitor, where the catalog is public, sees the configured instances and the hosted models every organization reaches, at the deployment's rates, and nothing that belongs to a tenant.
Enter your API key here (sent as Otari-Key header).
Compare prices for a request of this many input tokens: each model's minimum is taken from the pricing tier that request would settle at. Omitted, the base rates compare.
Case-insensitive text in a model's name, vendor, id, selectors, or provider instances.
For a signed-in caller, list only what a request billed to this workspace may call, with its own disabled keys and model lists applied. A workspace the caller cannot see answers 404. Ignored for an API key, which already names its workspace.
Number of matching models to skip.
0Maximum number of models to return.
100Match any named provider instance.
Match any vendor; an empty value names unknown vendors.
Require every input modality.
Require every output modality.
Minimum context window; unknown windows do not match.
0Maximum cheapest input price per million tokens; unpriced models, and models priced in another unit, do not match.
allPossible values: allPossible values: Only models at least one provider declared to keep no prompt or completion data offers.
falseRelease window ending today (UTC); zero disables it. Unknown and future releases do not match.
0namePossible values: ascPossible values: Include the filter choices drawn from the whole authorized catalog.
falseSuccessful Response
The grouped catalog, and the facts a reader needs to interpret its prices.
Models matching all filters before paging.
Whether an unpriced model is metered at the models.dev default.
When the accepted models.dev snapshot, which the defaults are billed from, was taken. Null while the bundled snapshot serves.
False when models.dev could not be read; descriptions are then absent.
Validation Error
GET /api/v1/catalog/models HTTP/1.1
Otari-Key: YOUR_API_KEY
Accept: */*
{
"count": 1,
"default_pricing": true,
"defaults_as_of": "2026-01-01T00:00:00.000Z",
"facets": {
"providers": [
"text"
],
"total_count": 1,
"types": [
"chat"
],
"vendors": [
{
"value": "text",
"vendor_slug": "text"
}
]
},
"metadata_available": true,
"models": [
{
"capabilities": {
"attachment": false,
"reasoning": false,
"structured_output": false,
"temperature": false,
"tool_call": false
},
"context_window": 1,
"deprecated": false,
"description": "text",
"discovered": true,
"family": "text",
"id": "text",
"input_modalities": [
"text"
],
"knowledge_cutoff": "text",
"max_output_tokens": 1,
"min_input_price_per_million": 1,
"min_output_price_per_million": 1,
"name": "text",
"offering_count": 1,
"open_weights": false,
"output_modalities": [
"text"
],
"price_sources": [
"organization"
],
"price_unit": "token",
"provider_count": 1,
"providers": [
"text"
],
"release_date": "text",
"resolves_to": "text",
"selector": "text",
"selectors": [
"text"
],
"type": "chat",
"unpriced_count": 1,
"vendor": "text",
"zero_data_retention_providers": [
"text"
]
}
]
}One model and every offering of it this caller may use.
The whole merged catalog is built and grouped to answer for one model. That is deliberate: the identity a model is found by is a property of the group, so narrowing the build to one model would need the grouping done first. The query count is constant; the cost is CPU per page view, growing with the size of the catalog rather than with the number of readers.
A model the caller may not see answers 404, the same as one that does not exist, so the route cannot be used to probe the catalog behind an allow-list. A signed-in caller's offerings also carry their organization's own usage of each over the last 30 days.
Enter your API key here (sent as Otari-Key header).
For a signed-in caller, only the offerings a request billed to this workspace may call.
Successful Response
One model with everything the detail page shows.
The largest any offering serves.
Whether an unpriced model is metered at the models.dev default.
True only when every offering with metadata says so.
falsemodels.dev's, from the offering that named the model.
Whether any offering was discovered from its provider.
The catalog id, vendor-qualified where the vendor is known: z-ai/glm-5.3, else the bare slug.
The largest any offering serves.
The cheapest offering's, per million of price_unit, at the comparison context where one was asked for.
falseWhat the cheapest prices are per: the unit every priced offering is priced in. Null where no offering is priced, or where offerings are priced in different units, which leaves both null.
The provider instances offering it, sorted.
The offering selector resolves to.
The id as a selector: send it as model and the best offering the caller can reach answers. Priced offerings come first, and among them the organization's own key, then the vendor's own provider, then the cheapest. Null until the gateway has indexed the catalog.
Every offering's selector, so the list can be searched by one.
What kind of model it is: the kind most of its offerings give, from the provider's listing or models.dev. A tie goes to the vendor's own provider. Null where no offering gives one.
How many offerings carry no price for this caller.
The provider instances offering it that are declared to keep no prompt or completion data, sorted.
Validation Error
GET /api/v1/catalog/models/{model_id} HTTP/1.1
Otari-Key: YOUR_API_KEY
Accept: */*
{
"also_available_from": [
{
"name": "text",
"provider_type": "text"
}
],
"capabilities": {
"attachment": false,
"reasoning": false,
"structured_output": false,
"temperature": false,
"tool_call": false
},
"context_window": 1,
"default_pricing": true,
"deprecated": false,
"description": "text",
"discovered": true,
"family": "text",
"id": "text",
"input_modalities": [
"text"
],
"knowledge_cutoff": "text",
"max_output_tokens": 1,
"min_input_price_per_million": 1,
"min_output_price_per_million": 1,
"name": "text",
"offering_count": 1,
"offerings": [
{
"context_window": 1,
"credential": "deployment",
"discovered": true,
"endpoints": [
"/chat/completions"
],
"max_output_tokens": 1,
"metadata_input_price_per_million": 1,
"metadata_output_price_per_million": 1,
"price_reference": "text",
"price_source": "organization",
"pricing": {
"cache_read_price_per_million": 1,
"cache_write_price_per_million": 1,
"input_price_per_million": 1,
"output_price_per_million": 1,
"pricing_tiers": [
{
"cache_read_price_per_million": 1,
"cache_write_price_per_million": 1,
"input_price_per_million": 1,
"min_input_tokens": 1,
"output_price_per_million": 1
}
],
"rates": {
"ANY_ADDITIONAL_PROPERTY": "anything"
},
"unit": "token"
},
"provider": "text",
"provider_type": "text",
"quantization": "text",
"selector": "text",
"short_selector": "text",
"usage_30d": {
"cache_hit_rate": 1,
"cache_read_tokens": 1,
"effective_price_per_million": 1,
"requests": 1,
"spend_usd": 1,
"total_tokens": 1
},
"zero_data_retention": false
}
],
"open_weights": false,
"output_modalities": [
"text"
],
"price_sources": [
"organization"
],
"price_unit": "token",
"provider_count": 1,
"providers": [
"text"
],
"release_date": "text",
"resolves_to": "text",
"selector": "text",
"selectors": [
"text"
],
"type": "chat",
"unpriced_count": 1,
"vendor": "text",
"zero_data_retention_providers": [
"text"
]
}Re-index the short spellings now, rather than on the refresher's next tick.
Enter your API key here (sent as Otari-Key header).
Successful Response
What the rebuilt index knows.
Slugs that resolve to an offering.
Selectors the deployment serves.
Pinned spellings, one per instance a model is offered on.
POST /api/v1/catalog/selectors/refresh HTTP/1.1
Otari-Key: YOUR_API_KEY
Accept: */*
Successful Response
{
"models": 1,
"offerings": 1,
"pinned_selectors": 1
}Last updated
Was this helpful?
