Checking whether token caching works on OpenRouter
is being used through to reduce token use and lower cost. is meant to save repeated input so the model does not have to process the same text again each time.
In practice, very little seems to be getting cached, so the savings look small. The main issue is how to set up correctly and how to check whether it is actually working.
Key points
- is being used with .
- The goal is to save tokens and reduce cost through .
- Very little input appears to be cached in the current setup.
- For , stable are important if caching is expected to save money.