LLM Cost Estimator
Estimate multi-model API cost from tokens or pasted prompts.
OpenPrice one prompt on five models and copy a tab-separated table into any spreadsheet.
Input tokens are estimated from its length.
Enter the typical length of a response.
Read the per-call cost for each model id.
Copy it or download a TSV for your spreadsheet.
You need a quick per-request number to drop into a spreadsheet, not a full monthly projection. The prompt is ready and you know roughly how long answers run. Cost Compare turns those two inputs into a five-row table.
Paste the prompt and set the expected number of output tokens (500 by default). Input tokens are estimated at about four characters each. The output is a tab-separated table with four columns: model id, input tokens, output tokens and the estimated cost in US dollars to six decimal places. It covers five models: gpt-4o, gpt-4o-mini, claude-sonnet, claude-haiku and gemini-flash, each priced at built-in reference rates.
Because the output is tab-separated, it pastes cleanly into Google Sheets or Excel, where you can multiply by call volume or replace the rates with your contract prices. You can also download it as cost-compare.tsv. Note that the rates here are separate from those in the LLM Cost Estimator, which adds daily and monthly projections across seven models, so do not mix numbers from both in one budget. The LLM token cost guide explains what list prices leave out.
The estimate is computed in the page from your text length and the number you enter. The prompt itself is not sent to any of the providers.
Keyboard shortcut
Answers for this tool. For site-wide help, open the FAQ hub.
Cost Compare prices a single call on five models and outputs a plain tab-separated table. LLM Cost Estimator covers seven models, takes calls per day and days per month, and produces a monthly projection with CSV export.
Run the prompt a few times and look at typical answer length, then divide characters by four. If you set a max tokens limit on the request, using that number gives you a worst-case figure.
They are fixed reference values in the tool and can lag behind vendor changes. Confirm against the provider pricing page before you quote numbers to anyone.