Credits and billing
Trial credits, pay as you go under a spend cap, and what each outcome is charged.
Base Eleven's models are fictional and its API returns simulated responses. Usage, credits and billing run for real, in test mode.
Free trial credits
A new workspace gets $5 in credits for 14 days, no card needed. Every call’s list price is debited from them when it ends. A typical call, 300 tokens in and 400 out, costs $0.0046 on Eleven Lite.
Before a call runs, its worst case must fit the remaining credits: the input at the input rate plus max_tokens at the output rate. If it does not, the call is refused with 402 credits_exhausted and nothing is charged. A smaller max_tokens fits more calls into the last of the credits.
When the 14 days are up, calls are refused with 403 trial_expired until a card is added or a commit is signed. Unused trial credits are not carried over.
Pay as you go
Add a card and the workspace moves to Pay as you go. Nothing is prepaid: usage is billed monthly, in arrears, at list prices, and the card is charged for the month that just ended. Tiers 2 to 5 apply.
A monthly spend cap protects you. It is $100 unless you set it lower on the settings page. Before a call runs, its worst case (input at the input rate plus max_tokens at the output rate) must fit under the cap for the month; if it does not, the call is refused with 402 spend_cap_reached and nothing is charged. The cap resets on the first of the month, UTC.
{
"type": "error",
"error": {
"type": "billing_error",
"message": "This call could take the workspace past its monthly spend cap.",
"b11_code": "spend_cap_reached"
},
"request_id": "req_01M3VNBWG00W3GE1R70W3GE1R7"
}What a call is charged
| Outcome | Charged |
|---|---|
| Completed (200, any stop_reason) | The input and output tokens in usage. |
| Refused (400, 401, 402, 403, 404, 409, 413, 429, 503) | Nothing. |
| Failed before any output (500, 502) | Nothing. |
Stream failed after it began (the error event) | The output tokens already sent. The input is not charged. |
| Connection dropped mid-stream | The input and the output tokens sent before you left. |
Only the tokens actually used are charged; the worst-case reservation made before a call is released when it ends. Prompts and replies are never stored, only their token counts.
List prices
| Model | Input | Output |
|---|---|---|
| Eleven Lite | $2.00 | $10.00 |
| Eleven Max | $4.00 | $20.00 |
The same prices come back from GET /models as pricing.input_usd_per_mtok and pricing.output_usd_per_mtok. A call is priced per direction, tokens times the rate over one million, rounded half up to the nearest millionth of a dollar, then summed. Enterprise commits buy credits below list; see Enterprise.
Where to see it
The usage page shows the balance or the month’s spend against the cap, and tokens and list cost by day and by model. It catches up within minutes of a call. There is no balance or usage endpoint in the API yet; a request to one gets 404.
Next: Models