Hello! These last few weeks, I am noticing that models provided through groq ( Llama Maverick, Gemma, Mistral etc.) do not have consistent pricing per minute as indicated by Vapi. It sometimes exceeds the per minute price by 6 to seven times that amount. Can you please look into it? And no, the transcribers and voices are not the issue.

Same for me…

A response would be greatly appreciated! @User

Could you please provide a call id? We would like to see how many prompt and completion tokens are being used for the call

Here is the call ID for this specific call: 019aa1a6-6c08-7333-af2e-e9dcefb06c35

Thanks for the call ids! We reviewed the logs and saw a potential issue. Because the characters are not typical unicode and the size of the prompt is fairly large, its causing the call to become very expensive. We recommend slimming down the prompt size and potentially switching the prompt to English then adding a single line to the prompt saying "only communicate in Greek". You can also test this by using an extremely simple prompt for your assistants in your squad to review the call cost and compare them to your previous calls.

Thank you for the reaponse! I will let you know if it gets fixed