Skip to content
Basevex AI is in private preview. These docs describe the V1 product; some features are still rolling out to design partners.

Usage & billing

What you pay

A monthly plan fee plus input and output tokens at the rates for your base model, shown per million tokens on the pricing page. Tokens are billed in addition to the plan fee. Payment is by card through Stripe; invoices are issued monthly.

The sandbox

Organizations without an active subscription are in the sandbox. You can use the console and call the API, but every response is synthetic, labeled with x-basevex-synthetic: true, and never billed. Sandbox usage appears in your dashboard separately from billable usage.

How tokens are counted

  • Billable tokens are exactly the prompt_tokens and completion_tokens the inference runtime reports. Counts sent by clients are ignored.
  • Requests that fail before producing any output are recorded with zero billable tokens.
  • If you cancel a stream, tokens the runtime reports as already processed are billed.
  • If the runtime can’t report usage for a cancelled stream, we record an estimate for your reference but do not bill it.
  • Each request is recorded once, keyed by its x-request-id. Retries of a failed request are separate requests.

Spend caps

Owners and billing managers can set a monthly spend cap (calendar month, UTC). To make the cap hold even with many requests in flight, Basevex reserves the maximum possible cost of each request before running it:

  1. Reservation = estimated input tokens × input rate + max_tokens × output rate.
  2. If spend so far plus all open reservations plus this one would exceed the cap, the request is rejected with 402 spend_cap_reached.
  3. When the request finishes, the reservation is replaced by the actual cost.

Setting a realistic max_tokens lets more requests through close to the cap. Owners and billing managers receive email at 80% and 100% of the cap.

Per-client caps

Each client can have its own monthly cap. A request attributed to a capped client must fit within both the client’s cap and the organization’s cap; if either would be exceeded, it’s refused with 402 spend_cap_reached.

Subscription status

StatusInference
No subscription, or canceledSandbox: synthetic responses, never billed
activeReal inference, billed
past_due, unpaid402 subscription_inactive until the payment method is updated

Invoices and estimates

Month-to-date amounts in the dashboard are estimates. Your invoice is issued by Stripe from the same usage records. Corrections, if ever needed, appear as separate adjustment lines — recorded usage is never edited.