How API pricing works
AI APIs charge per token, a piece of a word. In English, a token is roughly three quarters of a word, so 1,000 words is about 1,300 tokens. Every request pays for the input tokens you send (the instructions, history and documents) and the output tokens the model writes back. Output usually costs several times more than input.
Prompt caching cuts the cost of input you send again and again, such as a long system prompt: cached input is billed at a small fraction of the normal price. The LLM API cost calculator compares every model for your own request size and volume.
Price range by provider
Standard prices per million tokens (input / output), checked November 3, 2026:
| Provider | Lowest-priced model | Highest-priced model |
|---|---|---|
| Anthropic | Claude Haiku 5.5: $0.1 / $0.5 | Claude Fable 5.1: $10 / $50 |
| OpenAI | GPT-6 Luna: $0.1 / $0.5 | GPT-6 Astra: $10 / $50 |
| Gemini 2.5 Flash-Lite: $0.1 / $0.4 | Gemini 3.1 Pro Preview: $2 / $12 |
Questions
How many tokens is a word?
In English, about 0.75 words per token, or roughly 1,300 tokens for 1,000 words. Other languages and code use more tokens per word.
Why does output cost more than input?
Generating tokens takes more computing work than reading them, so providers price output tokens several times higher than input tokens.
What is prompt caching?
When the start of your prompt repeats across requests, providers can reuse it and bill those cached input tokens at a much lower rate.
How current are the prices?
They were checked against each provider's official pricing page on November 3, 2026. Providers change prices, so check their page before you commit.
More calculators
Tax calculators · Health calculators · Student loans calculators · Mortgage calculators · Money calculators · All tools