1
Audit Your Model Usage
Check the dashboard to see which models consume the most credits. Identify requests that could use a cheaper model.
2
Tier Your Requests
Route complex tasks to powerful models and simple tasks to cheap or free models. One API, different model IDs.
3
Apply :floor for Batch Work
Use the :floor variant for any background processing, data extraction, or bulk classification where speed is not important.
4
Consider BYOK for Heavy Providers
If most of your usage goes to one provider, BYOK gives you 1M free requests per month, with only a 5% fee after that.
How pricing works
Each model on OpenRouter has its own per token price, set by the providers who host it. OpenRouter charges a 5.5% platform fee ($0.80 minimum) when you purchase credits with a credit card (5% for crypto). There is no markup on inference pricing itself. The models page shows the exact input and output price per million tokens for every model, making it easy to estimate costs before committing.