AI API Cost Calculator
Estimate workflow cost before routing production requests through MetisRouter. Use published model pricing and bounded parameters.
Last verified: 2026-06-05
Configuration summary
How to estimate token model cost
Use the model catalog input and output prices, then multiply by expected input tokens and max output tokens. For low-balance accounts, set max_tokens so the platform can reserve bounded risk before upstream routing.
- Estimate prompt, context, tool output, and response size.
- Use separate calculations for cache read/write branches when the model publishes them.
- Check actual billed usage in Usage Logs after the first test request.
How to estimate image, video, and audio cost
Creative and media models may bill by image, second, duration, resolution, quality, input type, or task. The request body should include price-driving fields explicitly.
- Images: count, size, quality, edit/reference input.
- Video: seconds, duration, resolution, text-to-video vs image/video input.
- Audio and transcription: generated or processed audio duration and model-specific unit.
How to estimate recurring workflow cost
For n8n, Make, Zapier, Pipedream, LangChain, and SaaS jobs, estimate one successful run, then multiply by daily or monthly run count.
- Use low-cost models for high-volume extraction and classification.
- Use stronger reasoning models only for harder branches.
- Create separate API keys per workflow to keep reporting clear.
FAQ
Does this calculator replace the model catalog?
No. Use it as a planning guide, then confirm exact pricing in /models and actual billed usage in Usage Logs.
Why do I need max_tokens?
Bounded output lets MetisRouter reserve a maximum cost before upstream routing, especially when account balance is low.