Calculator

Fine-Tuning ROI Calculator

Model the training bill, shorter prompts, cheaper serving model, and break-even month for a fine-tuning project.

Fine-tune economics

Baseline model

Served / tuned model

Requests / day6,500

Baseline input tokens4,200

Baseline output tokens850

Tuned input tokens1,600

Tuned output tokens650

Tuned inference premium

Break-even

0.0 months

$64.00 upfront training cost

Before

$4,943

Claude Sonnet 4.6

After

,357

GPT-5.4 mini

Monthly savings

$3,587

73%

Option	Rates	Monthly
Baseline run-rate 4K in / 850 out	$3.00 / 5.00	$4,943
Tuned run-rate 2K in / 650 out	$0.750 / $4.50	,357

Fine-tuned inference is usually billed above the base list rate (the tuned-inference premium, default 1.5×) and a shorter prompt is assumed (fewer few-shot examples). Training rate and premium are editable assumptions — plug in your exact vendor quote.