Fine-Tuning Cost Calculator

Fine-tuning is billed per token trained. One pass over your dataset is examples × average tokens per example; multiply by the number of epochs to get the total trained tokens, then divide by one million and multiply by your per-1M-token training rate.

No. The small, mid, and large tiers only pre-fill an example training rate into the editable price field. Always replace it with your provider's current published fine-tuning rate before trusting the estimate.

Add the tokens in a typical prompt and its completion, then average across your samples. As a rough guide, one token is about four characters of English text, so a 2,000-character example is roughly 500 tokens.

No. This estimates only the one-time training cost. Running the fine-tuned model afterwards is billed separately per request — use an LLM token cost calculator for that part.