I’m checking the logs and I noticed Vapi is using fine-tuning model for some requests. This doesn’t make sense to me because I set gpt-4.1-nano as the default model. I understood that under the hood Vapi may switch models depending on the complexity of the situation. However, the logs show fine-tuning models being used, for example: - gpt-4.1-mini-2025-04-14 - gpt-4.1-nano-2025-04-14 Does anyone know why this happens? This makes me spend too much money

I'm not sure your question is just about pricing but on my end I use my own API keys for the model. VAPI says their pricing is pass-through but it's not. Compare prices between VAPI with their model and VAPI with your own keys, you'll see.

My concern is why it isn’t possible to set a model like gpt-4.1-nano as the default and prevent it from using gpt-4.1 (which is more expensive) behind the scenes. Also, it should never use fine-tuned models like gpt-4.1-mini-2025-04-14.

This should not normally occur unless a fallback model has been configured and triggered. Could you please share the call ID where this happened? I’ll review the logs to investigate further.