BYOK pricing and cost accounting

When your organization routes requests using Bring Your Own Key (BYOK) credentials, upstream model providers bill your accounts directly while Zorveus provides gateway routing, spend caps, and attribution.

Monthly pricing tiers

Zorveus includes a free monthly routing allowance for all BYOK traffic:

The three financial amounts in BYOK

Every BYOK inference call generates three distinct financial amounts:

1. Virtual spend (virtual_spend)

• Reference-priced value of AI usage consumed against spending caps and product-user credits.
• Calculated with the exact same reference pricing as wallet-funded requests.
• Changing from wallet to BYOK does not alter how much product allowance your customers consume.

2. Sell cost (sell_cost)

• Amount deducted from your Zorveus organization wallet.
• Under the free monthly threshold, sell_cost is $0.000000000000.
• For volume exceeding the threshold, sell_cost reflects the configured platform routing fee.

3. Provider cost (provider_cost)

• Upstream wholesale cost charged by the model provider (OpenAI, Anthropic, Google, etc.).
• The provider bills your direct credit card or invoice. Zorveus does not pay the provider for BYOK calls.
• Recorded in Zorveus for cost reporting and monthly fee calculation.

Sample BYOK usage event

Here is how a BYOK usage record appears in webhook events and API responses:

{
  "usage_event_id": "ue_01J8N2K3P4Q5R6S7T8U9V0W1X2",
  "billing_mode": "byok_external",
  "provider": "openai",
  "model": "openai/gpt-4.1-mini",
  "input_tokens": 1000,
  "output_tokens": 150,
  "cache_read_input_tokens": 0,
  "cache_creation_input_tokens": 0,
  "virtual_spend": "0.000350000000",
  "provider_cost": "0.000210000000",
  "sell_cost": "0.000000000000",
  "currency": "USD",
  "external_user_id": "usr_customer_8492",
  "status": "succeeded"
}

Platform fee calculation and wallet funding

When monthly BYOK provider cost crosses the $25,000 threshold, Zorveus calculates the platform fee only on the excess:

feeable provider cost =
    max(0, monthly provider cost after request - threshold)
    - max(0, monthly provider cost before request - threshold)

platform fee = feeable provider cost * fee percentage

The organization wallet pays this fee. Product-user credits never pay platform fees.

Before dispatching a feeable BYOK request, Zorveus places an estimated fee reservation on the organization wallet. If the wallet lacks funds, the request returns HTTP 402 Payment Required (zorveus_reservation_insufficient_balance). Maintain a funded wallet balance if you expect monthly BYOK traffic above $25,000.

Automatic fallback to wallet

Organizations can enable auto_fallback_to_wallet on their BYOK routing policy:

1. Routing order: Zorveus tries eligible BYOK credentials first. If all BYOK credentials encounter provider rate limits (429), quota exhaustion, or timeouts, the gateway routes to Zorveus-managed wallet routes.
2. Fallback event: When fallback succeeds, the usage event records billing_mode: "wallet" and logs the failed BYOK attempts in the provider attempt history.
3. No fallback on user errors: Fallback never triggers on invalid prompts, context length errors, or bad parameters (HTTP 400).