How model usage is billed
Who pays for a model request depends on the credential it runs on. On Alfe’s pooled access, it’s billed through Alfe. On your own key or subscription, your provider bills you and Alfe deducts nothing.
The three sources
Section titled “The three sources”Every model request falls into one of these sources, which match the labels on the Usage page:
| Source | Who bills you | What you see in Alfe |
|---|---|---|
| Alfe credits | Alfe | The billed amount |
| Own API key | Your provider | An estimate, not billed by Alfe |
| Subscription | Your provider | An estimate, not billed by Alfe |
How pooled usage is charged
Section titled “How pooled usage is charged”Alfe measures each request by the tokens it uses, applies that model’s rate, and converts the result to US dollars. Stronger models generally cost more, which is why choosing a default model per agent matters.
Pooled usage then follows the same order as your other usage:
- Your plan’s included allowance covers it, if your plan has one.
- Otherwise, it draws from your prepaid credit pool.
If neither can cover a request, Alfe refuses the request rather than charging you unexpectedly.
Keep an eye on spend
Section titled “Keep an eye on spend”These tools help you stay in control of model spend:
- Usage page — break spend down by provider, product, agent, team, or project.
- Usage limits — cap spend for the workspace, or narrow a cap to an agent, team, project, or service. See Usage and top-ups.
- Model policy — restrict models to the ones you’ve approved.