Models & Routing
Limits & Credits
What stops a request, and what it costs.
Credit balance
Paid models need a positive balance. At zero or below, a request to a paid model answers 402 Insufficient credits before anything is sent to a carrier. Free models keep working at any balance. Add credits in Settings → Credits.
How a request is metered
| Source | Used when |
|---|---|
Carrier-reported usage.cost | The carrier returns its own cost for the request |
| Catalog price × reported tokens | Otherwise: prompt tokens × prompt price + completion tokens × completion price |
| $0 | Free models, and requests that fail at the carrier |
The cost is deducted once, when the response completes. A stream you cancel midway is settled at the usage reported up to that point.
Checking credits
total_credits is everything you have added; total_usage is everything you have spent. Your balance is the difference.
shell
curl https://reroute.wtf/api/v1/credits \
-H "Authorization: Bearer $REROUTE_API_KEY"
{
"data": { "total_credits": 25, "total_usage": 1.0342 }
}Rate limits
Reroute does not add its own request-rate limit on top of your balance and key limits. Carriers do have limits; when a carrier rate-limits a request you receive 429 with the carrier's message, and can retry with backoff.