Pricing
AIVAX charges usage to the account balance in USD. Each product documents its own rates in a dedicated pricing section, linked below. This page covers the rules shared by every product: subscription allowances, storage, the request cost header, the data sharing discount, plan multipliers, credit top-up fees, and taxes on payments.
See subscription pricing for monthly plan prices and Plans and limits for account-level quotas. M means one million tokens; 1k means one thousand units. Approximate prices (~) vary with the model used and the work performed.
Plan list prices #
Use the pricing page when you are choosing a plan in a browser. To read the same public list prices from an application, call GET /api/v1/information/plans.json. No API key is required. The response lists Free, Pro, Max, and Ultimate. Each entry has plan and monthlyPrice in USD. Free is 0 because that plan has no subscription fee. A negotiated Custom price, or the price currently applied to an account, is not included. Use the Subscriptions API when you need the authenticated account’s price.
Product pricing #
| Area | Where to find rates |
|---|---|
| Inference and moderation | Inference |
| Agentic Tests | Agentic Tests |
| Voice and Realtime | Voice and Realtime |
| RAG collections, search, and answers | Collections |
| Media Injector | Documents |
| Reflex | Reflex |
| Web search and X search | Web Search |
| Fetch and OCR | Fetch and OCR |
| Image generation | Image generation |
| Text-to-speech and speech-to-text | Text-to-speech and Speech-to-text |
| Media descriptions, decisions, classification, and segmentation | Data processing pages |
| Document extraction and summarization | Extraction and Summarization |
| Translations | Translations |
| Built-in tools and memory | Built-in tools |
Included daily allowances #
Every plan includes separate daily allowances for eligible RAG embeddings, Reflex reranking, eligible semantic decisions, and Fetch/OCR extraction. Product rates apply when a metered item is not covered. Coverage is all-or-nothing per item, not necessarily per complete request: an item that cannot fit within the remaining allowance and its permitted margin is billed in full. Compare the allowances and check exclusions in Plans and limits. These daily allowances do not cover text-model inference; see Subscription models. For what each allowance covers, see Free RAG API on AIVAX Free.
Subscription models #
Pro, Max, and Ultimate include a monthly subscription model quota whose window starts with the first request that uses it and renews one month later. It applies only to subscription models, the catalog models whose names start with @aivax-subscription/. Using them never debits the account balance and they have no per-token prices; other models are charged to the balance at their rates. Free and Custom plans have no quota, so requests to subscription models on those plans return 429 Too Many Requests, as they do when a paid plan’s quota is exhausted. An exhausted quota is never charged to the balance or routed to another model.
The usage included by each plan is published as an equivalent at the models’ integrated catalog prices, rounded up to the nearest $10. The consumption rate is dynamic and can change daily, so the same model can use slightly more or less of the quota over time. See Subscription models and Monthly subscription model quota.
Storage #
Each plan includes storage. Pro, Max, and Ultimate overages are billed hourly at the monthly rates below; Free storage cannot be expanded. Document text stored in a collection counts as UTF-8 bytes. Text with accents or non-Latin characters can use more of the included quota than basic Latin text of the same length.
| Description | Pricing |
|---|---|
| Free storage | 30 MB included; no expansion |
| Pro storage | 2 GB included; excess $0.50/GB/month |
| Max storage | 10 GB included; excess $0.30/GB/month |
| Ultimate storage | 20 GB included; excess $0.20/GB/month |
Request cost header #
API responses include a Consumed-Credits header when the request charged usage to your account balance. Its value is the total charged by that request in USD, written as a decimal number with a dot separator, for example Consumed-Credits: 0.000184. It already includes the plan multiplier and covers every item charged during the request, such as inference, embeddings, reranking, and tool usage.
Cross-origin responses expose Consumed-Credits and X-Request-ID, so browser JavaScript on another origin can read them when they are present. X-Request-ID is the identifier AIVAX assigned to that request.
The header is omitted when the request charged nothing, including requests rejected by validation or covered entirely by a daily subscription allowance or by the monthly subscription model quota. Treat a missing header as zero.
The value only covers charges recorded before the response is sent. Usage recorded later, such as asynchronous Batch work, is not included. Streamed responses never include this header, because it is added after the response has been produced. For a streamed request, read the cost recorded with the conversation (available when conversation logging is enabled) or calculate the cost from the reported token counts. Use your account usage records as the source of truth for billing.
Parse the value as a floating-point number. Very small amounts, such as 1E-05, can be written in scientific notation, and the sum of several charges can carry floating-point noise in its last digits.
Data Sharing discount #
Accounts that enable Data Sharing receive a 10% discount on eligible RAG query-embedding usage and eligible reranking operations performed while sharing is enabled. Other services keep their regular prices.
Plan multipliers #
Usage rates are subject to the plan’s inference tax:
| Plan | Inference tax | Daily ceiling on model inference |
|---|---|---|
| Free | +25% | $1.50 |
| Pro | +5% | $1.20 |
| Max | +2% | $1.00 |
| Ultimate | 0% | Not applicable |
For integrated-model inference, the tax is recorded as its own usage entry, inference.commission, under the model that served the request, so token entries show the model’s rates. Once the day’s inference.commission entries reach the ceiling, further model inference is charged at the model’s rates only. The day runs from midnight to midnight in the America/Sao_Paulo time zone, and the count starts again at midnight.
Other services subject to the plan multiplier include it in their own rates, and those charges don’t count toward the ceiling.
BYOK usage is not affected by inference taxes. Product pages state when a rate is exempt from the plan multiplier, as with image generation.
Credit top-ups #
Adding credits charges the amount you choose plus the payment processor fee and the taxes on payments. Only the amount you choose is added to your balance; fees and taxes are not. On Free, an AIVAX service fee of 5% of the amount, with a $0.20 minimum, is also added. Pro, Max, and Ultimate pay no AIVAX service fee. The dashboard shows each fee, the taxes, and the total before you confirm the payment.
Taxes on payments #
Every payment includes Brazilian taxes: credit top-ups, subscription payments and renewals, and bills issued by the AIVAX team. The taxes are added on top of the amount due, so the charged total is the amount divided by (1 - 0.1333). They correspond to approximately 11.33% of the total in service taxes and 2% in IOF.
For example, a $19.00 subscription payment charges $21.92, and adding $10.00 in credits on a plan with no service fee charges $11.54 before the payment processor fee. Taxes are not added to your balance, and listed plan prices do not include them. The receipt of each payment lists the amount and each tax.
