How current are the model prices?

They are reviewed against the providers' own pricing pages and updated when rates move. Every calculator shows the rate it used, so you can check it against the provider and override it if you are on custom pricing.

How accurate is the token count?

The counter uses the same tokenizer family the models use, so for English text it is exact rather than an estimate. Counts differ between model families, so pick the model before reading the number.

Is my prompt sent anywhere?

No. Counting and pricing happen in your browser, so the prompt you paste in never leaves your device. That is deliberate: the prompts worth pricing are usually the ones you cannot paste into someone else's site.

Do these include the cost of reasoning tokens?

The reasoning token calculator does, and it is the one to use for models that think before answering. The plain cost calculator prices input and output only, which will understate the bill for a reasoning model.

Yes, free with no account, paid for by light advertising.

AI cost calculators

Work out what an LLM feature will cost before you ship it, not after the invoice.

Open the tools 33 tools, all in your browser

API pricing is quoted per million tokens, which is a unit nobody thinks in. These calculators turn it into the numbers you actually plan against: what one conversation costs, and what that becomes at a thousand users a month.

Tokens are the thing that trips people up. A token is roughly four characters of English, so a 500-word email is about 650 tokens, and an image is priced by how many tokens it resolves to rather than by its file size. The token counter and the image token calculator show you the count for your own text and pictures instead of a rule of thumb.

Prices change often, so we check the model rates against the providers' own pricing pages, and every calculator shows its working. If a number looks wrong you can see which rate produced it and override it. Nothing you paste is uploaded: the maths runs in your browser, which matters when the prompt you are pricing is a real one from your product.

Which tool for which job

LLM cost calculator: Price a single call once you know roughly how long the prompt and the answer run.
AI chatbot cost calculator: Scale one conversation up to a monthly bill for a real number of users.
Token counter: Count the tokens in your own prompt instead of guessing from word count.
Context window planner: Check whether the documents you want to stuff into one call will actually fit.
Model switch savings calculator: See what you would save moving a workload from one model to a cheaper one.
Prompt caching savings calculator: Work out whether caching a long system prompt pays for itself at your call volume.
Self-host vs API calculator: Compare GPU hours against per-token pricing before committing to running your own.

Questions about the AI calculators

How current are the model prices?: They are reviewed against the providers' own pricing pages and updated when rates move. Every calculator shows the rate it used, so you can check it against the provider and override it if you are on custom pricing.
How accurate is the token count?: The counter uses the same tokenizer family the models use, so for English text it is exact rather than an estimate. Counts differ between model families, so pick the model before reading the number.
Is my prompt sent anywhere?: No. Counting and pricing happen in your browser, so the prompt you paste in never leaves your device. That is deliberate: the prompts worth pricing are usually the ones you cannot paste into someone else's site.
Do these include the cost of reasoning tokens?: The reasoning token calculator does, and it is the one to use for models that think before answering. The plain cost calculator prices input and output only, which will understate the bill for a reasoning model.
Are they free?: Yes, free with no account, paid for by light advertising.