AI API Cost Calculator
Estimate per-call, monthly and yearly costs for GPT, Claude, Gemini and other LLM APIs from token counts and call volume.
Usage inputs
Cost estimate
Estimated cost per month
1 token is about 0.75 English words, so 1,000 input tokens is roughly 750 words.
How to use the AI API cost calculator
Pick a model preset or enter your own prices per million tokens. Add your average input tokens per call (1,000 is about 750 words of context), your average output tokens (500 is about 375 words) and your expected monthly call volume. Cost per call, per month and per year update as you type.
Why model costs before you build
AI API costs scale with usage. A feature that costs $20 a month at 1,000 calls costs about $20,000 a month at a million calls, which can be the difference between a profitable feature and a painful bill. Modelling cost early helps you choose the right model. Smaller models such as GPT-4o mini or Gemini Flash cost a fraction of flagship models and handle many routine tasks well, so test them first.
Common use cases
- Forecasting monthly costs before launching an AI feature
- Comparing models for a specific use case
- Building a business case for an AI product investment
- Budgeting AI marketing automation
- Estimating savings from switching models
- Planning headroom for usage spikes
Token rules of thumb
1 token is about 0.75 English words or 4 characters. A typical email is around 200 tokens, a blog post draft around 2,000 and a 10-page document around 5,000. Other languages and code often use more tokens per character, sometimes 2 to 3 times more for Chinese, Japanese or Korean text.
Frequently asked questions
Are the preset prices current?
Presets are standard list prices from when this tool was built, and providers change pricing often. Check each provider's pricing page and type the current figures into the price fields before you commit to a budget.
What about caching and batch discounts?
OpenAI offers 50% off through its Batch API, and Anthropic offers prompt caching discounts of up to 90% on repeated context. This calculator uses standard rates, so lower the price fields to model those savings.
Why does output cost more than input?
Output tokens have to be generated one at a time, which takes more compute than reading input. Output usually costs 3 to 5 times more per token.
Should I use the cheapest model?
Test before you commit. Run a sample of real tasks through a small and a large model and compare quality. Save flagship models for tasks that need deep reasoning.