Pricing

AIVAX uses a prepaid account balance. Paid invoices add credit to the account, and usage records subtract from that balance.

The backend calculates balance as:

balance = paid, unexpired invoice total - usage total

Use the AIVAX pricing page for current commercial plan prices. This page documents billing behavior that is visible in the API source.

Credits and invoices

Credits are represented as invoices.

• Paid invoices increase the usable account balance until their expiration date.
• Unpaid payment invoices are created with a one-year expiration.
• Unpaid invoices older than three days are removed by cleanup.
• Expired paid invoices no longer count toward balance.
• Payment invoice creation requires at least 3 USD and is rate-limited.

Usage billing

Every billable operation writes one or more usage records. Each usage record has:

• Description.
• Unit price.
• Quantity.
• Optional model name.
• Usage category.
• Resources such as API key, gateway, or collection.

The final unit price is multiplied by the account tax multiplier and the current plan commission multiplier.

Plan	Commission multiplier
Free	1.25x
Pro	1.05x
Max	1.00x

Inference billing

Integrated model billing uses the model's pricing table from the backend. Pricing can vary by model and by input-token threshold. Usage can include:

• Text input tokens.
• Cached input tokens, when the selected model has cached-input pricing.
• Audio input tokens, when applicable.
• Output tokens.

Some integrated models can be covered by subscription-model reserve windows. When a model has a subscription usage multiplier and the account still has reserve remaining in both the six-hour and weekly windows, the recorded model usage can be covered instead of charged to balance.

Current reserve windows are:

Plan	Six-hour reserve	Weekly reserve
Free	None	None
Pro	250 units	3,000 units
Max	1,000 units	15,000 units

BYOK calls use your external provider key, but AIVAX still enforces BYOK request limits because the request passes through AIVAX infrastructure.

RAG billing

The default embedding model is @aivax/data-embedding. Its backend price is:

$0.015 per 1,000,000 input tokens

This applies to document indexing and query embedding usage that uses the default embedding model. Reranking, when enabled, can add separate reranking usage.

Storage billing

Storage is checked before balance-gated operations. If storage usage exceeds the plan quota, the request can fail with 402 Payment Required.

Plan	Included storage	Overage
Free	30 MB	No paid overage
Pro	2 GB	$0.50/GB/month, charged hourly
Max	20 GB	$0.20/GB/month, charged hourly

Storage overage is charged hourly. The job bills only excess usage above the included quota, and only when the billable excess is greater than 1 MB.

Storage can include RAG document content and vectors, long-term memory items, media description cache, web chat session content, and shell files.

Balance requirements

Billable routes check balance before running. The generic balance middleware rejects balances below the route minimum; chat clients, integrations, and batch processing also stop when the balance is zero or negative. Some multimodal chat-completion inputs require a minimum balance before the model call starts:

Input type	Minimum balance
Image or audio	$0.10
File or video	$0.50

If the account balance is too low, the API returns 402 Payment Required.

Plans and limits

Plans affect both price and operation:

• Model access.
• Commission multiplier.
• Request and token rate limits.
• BYOK request limits.
• RAG quotas.
• Tool limits.
• Storage quota and overage price.
• Conversation retention.
• Subscription-model reserve windows.

See Plans and limits for the technical quota matrix.