- Set the workload
Enter requests per month and any other volume assumptions that define the period or batch.
- Add usage assumptions
Complete the remaining usage fields, keeping token, duration, or size units consistent with their labels.
- Enter current rates
Use the current prices for the exact model, endpoint, quality, and processing tier you plan to use.
- Review the estimate
Read the main result and the supporting breakdown to see the largest cost or capacity driver.
- Test another scenario
Change one assumption at a time or select Reset to restore the default example values.
OpenAI Model Calculator
The OpenAI Model Calculator compares two text models on estimated workload cost, latency, and monthly spend. Both options use the same request volume and token assumptions so the difference comes from model-specific prices and response time.
Use it for early model selection, routing design, or cost-performance reviews. Quality is not reduced to a single score here; teams should pair this estimate with evaluations on their own tasks before choosing the lower-cost or faster option.
Calculation inputs
requests
tokens
tokens
USD/1M
USD/1M
seconds
USD/1M
USD/1M
seconds
Result
—
Estimated result
Model A monthly cost—
Model B monthly cost—
Cost difference vs A—
A minus B latency—
Model monthly cost = Requests × [(Input tokens × input rate + Output tokens × output rate) ÷ 1,000,000]. Savings = Model A cost − Model B cost.
The calculator applies the entered values directly and displays rounded results for planning.
What the result means
The main result shows the absolute monthly cost difference between the two entered model options.
Latency values are user-entered planning assumptions and should come from production measurements under comparable load.
Given: Example assumptions
Calculation: For 100,000 requests with 1,500 input and 350 output tokens, Model A at $2.50/$15 costs $900 per month. Model B at $1/$6 costs $360. Model B saves $540, or 60% relative to Model A, under the same workload assumptions.
Result: The displayed result follows the same formula and rounding logic.
Does a cheaper model always reduce total cost?
Not if it requires more retries, longer prompts, or additional calls. Adjust workload inputs to reflect those effects.
How should I compare model quality?
Run task-specific evaluations with representative examples and acceptance criteria; this calculator only compares cost and latency assumptions.
Can the two models have different token counts?
This version holds token counts constant. Run separate scenarios if one model consistently produces longer outputs or needs more context.
What latency should I enter?
Use median or percentile latency from the same endpoint, region, processing tier, and workload pattern.
Can I use this for model routing?
Yes. Compare candidate models, then estimate a blended route by weighting each model’s cost by the share of requests it handles.