Documentation
FAQ: Why did this model cost more than expected? in the NexoRouter documentation.
FAQ: Why did this model cost more than expected?
Model cost depends on the selected model, input tokens, output tokens, and the request pattern. Usage Logs are the source of truth for request-level cost.
Check in this order
- Confirm the exact model ID in Usage Logs.
- Check prompt and completion tokens.
- Check whether the app retried automatically.
- Check whether an agent loop made multiple requests.
- Check whether
max_tokensallowed a long completion. - Compare the model against Pricing.
Common causes
| Cause | Fix |
|---|---|
| More output tokens than expected | Set a smaller max_tokens during testing. |
| Agent loop or tool retries | Use a low-budget key for agents and inspect Usage Logs. |
| Expensive model selected by a tool | Restrict the key model scope or configure one model explicitly. |
| Automatic fallback chain | Start with one model only until cost is understood. |
| Long context prompt | Trim input, summarize history, or choose a model intentionally. |
Cost-control setup
| Environment | Recommended key setup |
|---|---|
| Local experiment | Low budget, one or a few low-cost models. |
| Staging | Separate key, realistic model scope, moderate budget. |
| Production | Separate key, approved models only, monitored spend. |