Hi Retell team,
Our agent’s average token consumption jumped from ~2800 to ~3600 tokens per call recently, without any changes on our side. Same prompts, same nodes, same models, same conversation flow — nothing was modified.
This is urgent because we charge our customers a fixed rate based on expected costs. The jump from ~2800 to ~3600 tokens pushes our agent over the 3500 token threshold, which triggers surcharges from Retell. This directly eats into our margins and makes our pricing to clients unsustainable — we are now losing money on every call.
We did not budget for this increase because we did not cause it. Our agent config has not changed.
Thank you for reaching out to Retell AI Support. We’ve received your ticket and our team will respond within 8 hours.
Hello,
Can you provide some call IDs where you have encountered this issue?
Regards,
Retell Support
Hello It is in my dashboard that the token went up dradtically so every call since that my dash board went up: call_fc0e46d1c5795a82623c7bea5f9 call_8ba54aee7d12f6c83ad2c9bb262 do you also want my agent id?
Hello,
We will forward this to our engineering team for further investigation and will get back to you shortly.
Regards,
Retell Support Team
Hi, looking at our internal dashboard, this is the trend for average number of input tokens per call for the agent used in those calls (agent_876d903a11ce59b40f4a5d91e5):
There does seem to be a slight increase starting around Jan 19, but it’s not a significant jump compared to the previous usage. Also, please note that we our token surcharge is proportional to how far over 3500 you are. For a call with 3600 tokens, we will only bill around 3600/3500 ~ 1.028x the original rate of the LLM.
