Models

The public Models catalog lists publicly discoverable models, including context window where applicable and public prices with explicit units. For the base URL and supported HTTP paths, see API endpoint.

The versioned machine-readable catalog supports validated speech-to-text and text-to-speech entries with explicit USD rates by unit. These catalog types do not add model API paths. See API endpoint for supported paths.

The public Models page and Console model list show cards with each model's rates and units. When a speech-to-text model has both duration and token pricing, the duration rate and the complete token-price group (text input, audio input, and text output) are alternatives, not additive charges. Cards omit price items that do not apply and show zero-priced items as “Not charged” without a unit.

In the Console's Models page, the URL preserves your search and selected model. Reload or reopen the link to restore that view. Returning from a model's details to the list keeps your search.

Calling a model

For chat models, pass the model id in your request body, exactly as shown in the catalog:

{
  "model": "YOUR_MODEL_ID",
  "messages": [{ "role": "user", "content": "Hello" }]
}

Decision models use a different path and a different request shape — see Decision API.

Pricing

Prices are what you pay, billed against prepaid credits. Text-model input, cache, and output rates are per 1M tokens. Audio prices identify input or output and text or audio. Duration rates are per minute, character rates are per 1M characters, and token rates are per 1M tokens. These display conversions use catalog rates and do not change the billing basis.

Models with context price tiers show each input-token range and its prices in the catalog and model details. The input count includes cached tokens. The matching range sets the prices for the whole request: input, cached input, and output. Only the matching range applies, rather than adding the tiers together. Below the first threshold, base prices apply. Models without context price tiers keep their flat prices.

The machine-readable catalog includes these rates in pricing.contextPriceTiers. Each thresholdTokens is the first input-token count for that tier. Prices use USD decimal strings per million tokens.

Availability

Model status in the catalog reflects whether the model is currently callable. Unavailable models remain listed for reference but reject new traffic.

If an upstream confirms that a model no longer exists and no other upstream can serve it, Tokener.ai removes it from model directories. Temporary failures do not remove models from directories.

Access and discovery

Restricted models are available to authorized organizations, including organizations bound to an authorized Partner. Your organization's model directory can therefore differ from the public catalog. API key model restrictions still apply.

Hidden models are omitted from model directories, including GET /v1/models. An authorized key can call a hidden model directly using its model ID. Absence from a directory does not establish whether a known model ID is callable. Model-access changes take time to apply; saving a change does not guarantee immediate revocation. If model discovery returns 503, direct model calls remain available subject to their normal authorization and availability.