Pricing
Every model costs exactly what its vendor charges. There is no subscription and no per-seat fee: you buy prepaid credits, and usage is deducted per token at the official rate. The fee schedule below is the whole story.
Tokens are billed at the vendor list price for input, cached input, cache writes, output and reasoning. Nothing is added.
Added once when you top up, covering payment processing. Purchases from $5 to $10,000; credits never expire.
Route through your own vendor accounts. The vendor bills you for tokens; ElevenRouter charges this percentage of the official price for routing, failover and observability.
What a request costs
Example: 20,000 input tokens and 4,000 output tokens on a model priced at $3 / $15 per million.
| Input · 20,000 × $3/M | $0.0600 |
| Output · 4,000 × $15/M | $0.0600 |
| Charged to your balance | $0.1200 |
Cached input tokens are billed at the vendor's cache-read rate when the vendor reports them. Send X-ER-Metadata: enabled to receive this breakdown on every response.
Never surprised by spend
- Zero-completion insurance. A response with no output tokens and no finish reason costs nothing, even when the vendor bills us for the prompt.
- Budgets with forecasts. Daily, weekly, monthly or lifetime caps per organization, workspace or key, hard (block) or soft (alert), each with a projection of when the limit will be reached.
- Key spend limits. Every API key can carry its own lifetime or recurring limit; alerts fire at 50, 80, 90 and 100 %.
- Alerts where you work. Low balance, budget and key-limit alerts by email, signed webhook or Slack.
- Receipts and invoices. Every purchase gets a receipt and a downloadable invoice with your legal name and tax id.
- Free credits first. Promo and granted credits are always spent before the credits you paid for.
Questions
Do you mark up model prices?+
No. Model usage is billed at the official vendor price. Our revenue comes from the fee on credit purchases and from BYOK routing.
Which vendor am I paying for a request?+
Responses name the model's vendor (for example Anthropic for Claude models) in the x-er-provider header and in the activity log. Prices are the vendor's published rates for that model.
What happens when my balance runs out?+
Requests fail with 402 insufficient_credits. Enable auto top-up on the Credits page to buy a fixed amount whenever the balance drops below a threshold, or set a low-balance alert.
Are failed requests billed?+
No. Requests that fail before producing output are never charged, and a response with zero completion tokens that ended without a finish reason is covered by zero-completion insurance. Partial streamed output that reached you is billed for the tokens you received.
Can I get a refund?+
Purchased credits that have not been used can be refunded to the original payment method; promotional and granted credits are not refundable. Contact support from the dashboard.
Do credits expire?+
Purchased credits never expire.
Is there a rate limit?+
Keys default to 600 requests per minute, adjustable per key. Responses carry x-ratelimit-limit and x-ratelimit-remaining; exceeding the limit returns 429 with retry-after.