Could someone help me understand the billing structure for finetuning gpt 4o-mini? When I go to start a finetune job, the job fails because it estimates a cost of around $40 which exceeds the limit I have set. In the documentation, it says that finetuning gpt-4o mini is free until September 2024 unless you exceed a certain number of tokens. Does this mean I am exceeding the token limit?
The daily free limit for training gpt-4o-mini is set at 2M tokens which is the equivalent of $6 as per the pricing for gpt-4o-mini fine-tuning. Any overage is charged at $3 per 1M training tokens in line with the normal fine-tuning pricing for the model .
So yes, if your cost is around $40, then you are exceeding the free token limit by more than 13M tokens.
1 Like