Set Model Price Override

PUT /model_pricing

Set Model Price Override

Set this organization's price for one (provider, model) pair, creating the override or replacing the existing one.

Every rate is replaced, not merged: a rate you omit is stored as 0, which means the token class bills at the prompt or completion rate rather than keeping whatever it was before. Send the full set you want in effect.

The new price applies to telemetry arriving after the write — already-recorded costs are not recomputed — and is republished to the gateways immediately so least_cost routing ranks on it.

Parameters

Request body

Content type: application/json

• provider (string, required):
• model_name (string, required):
• prompt_cents_per_1m (integer, optional):
• completion_cents_per_1m (integer, optional):
• cache_read_cents_per_1m (integer, optional):
• cache_write_cents_per_1m (integer, optional):
• reasoning_cents_per_1m (integer, optional):
• duration_cents_per_1m_seconds (integer, optional):

Responses

• 200: Stored

• model_pricing (object, required): The price one (provider, model) pair bills at for this organization. All token rates are integer cents per 1,000,000 tokens.

• model_pricing.id (string, required): Price row id. Pass it to the delete endpoint to drop an override; catalog rows cannot be deleted.

• model_pricing.org_id (string, optional): Organization that owns this override. Omitted on catalog rows, which belong to no organization.

• model_pricing.provider (string, required):

• model_pricing.model_name (string, required):

• model_pricing.prompt_cents_per_1m (integer, required): Input token rate.

• model_pricing.completion_cents_per_1m (integer, required): Output token rate.

• model_pricing.cache_read_cents_per_1m (integer, required): Rate for tokens read from an upstream prompt cache. 0 means no distinct rate — those tokens bill at the prompt rate.

• model_pricing.cache_write_cents_per_1m (integer, required): Rate for tokens written into an upstream prompt cache. 0 means no distinct rate.

• model_pricing.reasoning_cents_per_1m (integer, required): Rate for reasoning tokens. 0 means no distinct rate — they bill at the completion rate.

• model_pricing.duration_cents_per_1m_seconds (integer, required): Cents per 1,000,000 seconds, for models billed by audio length rather than tokens. 0 means not billed by duration. Operator-set only: the catalog carries no audio prices.

• model_pricing.source (string, required): Where the price came from. user is this organization's own override. The other two are catalog defaults and are mutually exclusive per deployment: models.dev when the control plane syncs the catalog online, snapshot when it seeds the catalog from a local file instead — which is what an air-gapped or offline installation does.

• model_pricing.effective_from (string, required):

• model_pricing.updated_at (string, required):

• 400: Request body or parameters failed validation.

• error (object, required): Error details.

• error.code (string, required): Machine-readable identifier. Examples: UNAUTHENTICATED, INVALID_REQUEST, MASTER_KEY_UNSET, INTERNAL.

• error.message (string, required): Human-readable explanation; safe to surface to operators.

• 401: Missing or invalid bearer. Response fields (application/json): identical to PUT /model_pricing, response 400 above.

• 403: Authenticated request is not permitted. Response fields (application/json): identical to PUT /model_pricing, response 400 above.

cURL

curl -X PUT '$AISIX_CP/model_pricing' -H 'Authorization: Bearer $AISIX_TOKEN'