AI API Cost Calculator
Compare estimated token costs for OpenAI, Claude, Gemini, and DeepSeek. Model a single request or forecast your full monthly AI bill.
Build your workload
Estimated at about 1.33 tokens per English word. Choose “Tokens” if you have exact usage data.
Choose models to compare
No models match this filter.
＋ Add a custom model or update a price
Custom prices are saved only in this browser.
Your cost comparison
Estimated for 3,000 monthly requests
Select at least one model
Your cost estimates and comparison chart will appear here.
Monthly cost at a glance
Lower is cheaper for this workload
Detailed comparison
Estimates use published token rates and your cache setting
| Model | Per request | Daily | Monthly | Annual | Cache saves |
|---|
How to calculate AI API costs
AI providers usually price text models per one million tokens. A token is a small piece of text: roughly four English characters or about three-quarters of a word, although the exact count varies by model and language.
((uncached input × input rate) + (cached input × cache rate) + (output × output rate)) ÷ 1,000,000 × monthly requestsThis calculator separates input, cached input, and output because output tokens are often more expensive. It then multiplies the per-request estimate by users, requests per user, and active days to create monthly and annual forecasts.
Why real API bills may differ
Tokenizers, reasoning tokens, tool calls, web search, image or audio input, retries, long-context pricing, regional endpoints, batch discounts, and provider-specific tiers can change your final bill. Use exact token usage from your API logs whenever possible and treat word-based calculations as planning estimates.
AI token cost calculator FAQ
How do I calculate the cost of an AI API request?＋
Multiply input tokens by the model’s input price per million, multiply output tokens by its output price per million, add both amounts, and divide by one million. Apply the lower cached-input rate only to eligible cache-hit tokens.
How many tokens are in 1,000 words?＋
A common planning estimate is about 1,333 tokens for 1,000 English words. The exact number depends on the model’s tokenizer, punctuation, formatting, code, and language, so exact API usage data is more reliable.
What are cached input tokens?＋
Prompt caching lets a provider reuse eligible prompt content instead of processing it at the full input rate. Providers have different eligibility rules, minimum prompt sizes, retention windows, cache-write charges, and cache-hit prices.
Does this include reasoning or thinking tokens?＋
Enter all billable output tokens reported by your provider, including reasoning or thinking tokens when the provider counts them as output. A word estimate cannot predict hidden reasoning usage precisely.
Are ChatGPT Plus or Claude Pro included?＋
No. Consumer chat subscriptions and developer APIs are billed separately. This tool estimates pay-as-you-go API token costs, not monthly app subscription fees.
Is the text I paste into this calculator private?＋
Yes. This version performs its estimate inside your browser and does not send pasted text to SaaSVerdicts or an AI provider. Avoid pasting sensitive information into any website unless you trust the site and device.
Estimates only. Prices exclude taxes, credits, free tiers, fine-tuning, storage, web search, tools, media, long-context premiums, and enterprise agreements unless you add them manually.