Pricing & credits

Every call debits credits based on the model and the options you choose (resolution, duration, etc.). You pre-fund a single balance — no seats, plans, or minimums; credits never expire.

• The exact price is returned as creditsCharged on every create call.
• Chat models bill per token and their non-streaming response carries usage: inputTokens, outputTokens, cachedInputTokens and cacheWriteTokens. The two cache counts are parts of inputTokens that were billed at the cached and cache-write rates; the model reports them, there is no field to request caching. Where a model's cached rate is actually different from its fresh input rate, the Pricing page and the model's own page list it as its own row.
• Per-model prices are shown on the Pricing page and each model's page.
• Failed media tasks are refunded automatically.
• Check your balance anytime via GET /api/v1/credits.