AI Token Cost Calculator
Estimate what a prompt or a whole workload will cost across Claude, GPT, and Gemini models. Enter token counts or paste text to estimate tokens, then compare per-model pricing side by side.
// 2,000 in + 500 out · 1,000 requests
| Model | Provider | Per request | Total | vs cheapest |
|---|---|---|---|---|
| Gemini 2.0 Flashcheapest | $0.00040 | $0.4000 | 1.0× | |
| GPT-4o mini | OpenAI | $0.00060 | $0.6000 | 1.5× |
| Gemini 2.5 Flash | $0.00185 | $1.85 | 4.6× | |
| o3-mini | OpenAI | $0.00440 | $4.40 | 11.0× |
| Claude Haiku 4.5 | Anthropic | $0.00450 | $4.50 | 11.3× |
| Gemini 2.5 Pro | $0.00750 | $7.50 | 18.8× | |
| GPT-4o | OpenAI | $0.0100 | $10.00 | 25.0× |
| Claude Sonnet 4.x | Anthropic | $0.0135 | $13.50 | 33.8× |
| Claude Opus 4.x | Anthropic | $0.0675 | $67.50 | 168.8× |
Estimates only. List prices as of January 2026; token estimates use ~4 characters per token. Confirm current pricing with each provider before committing spend.
About AI Token Cost Calculator
Large language models bill per token — roughly ¾ of a word — with separate rates for the tokens you send (input) and the tokens the model generates (output). This calculator lets you enter token counts or paste representative text, multiply by your expected request volume, and compare the total cost across Claude, GPT, and Gemini models so you can pick the right price/quality tier before you build.
Frequently asked questions
How accurate are these estimates?+
They use published list prices and a ~4-characters-per-token approximation. Real tokenization varies by model and language, so treat the numbers as planning estimates, not invoices.
What's the difference between input and output tokens?+
Input tokens are your prompt (plus any context you send); output tokens are the model's reply. Output is usually several times more expensive, so long generations dominate cost.
Why is one model so much cheaper?+
Smaller/faster models (Haiku, GPT-4o mini, Gemini Flash) are priced for volume; frontier models cost more per token but may need fewer retries. Compare on total cost for your actual workload.