How costs are calculated
Every request consumes prompt tokens (your input) and completion tokens (the model’s output). These are billed separately at the rates shown for each model.| Tokens | Price per 1M | Cost | |
|---|---|---|---|
| Prompt | 500 | $2.50 | $0.00125 |
| Completion | 200 | $10.00 | $0.00200 |
| Total | $0.00325 |
Markup
ModelSwitch adds a small markup over provider prices to cover infrastructure and support. The prices shown viaGET /v1/models already include this markup — there are no hidden fees.
Balance
Your balance must be positive to make requests. When your balance reaches zero, requests return an error until you top up. Your balance is stored in USD internally. It may be displayed in your local currency in the dashboard.How billing works under the hood
ModelSwitch uses pessimistic holds to prevent overdraft, particularly under heavy concurrent usage.Hold placed
When your request starts, an estimated cost is reserved (held) from your balance. Your available balance is reduced immediately.
Request forwarded
The request is forwarded to the upstream provider. If the request fails before completing, the hold is fully released and your balance is restored.
Managing your balance
Top up
Add funds to your balance via card payment
Auto top-up
Set up recurring top-ups so you never run out
Documents
Download invoices, acts, and UPD documents