AI cost calculator
AI token cost calculator
One workload, every current model. Enter how many requests you send and how many tokens each one uses, and see the monthly bill on Claude, GPT and Gemini at each provider's list price, side by side.
Your inputs
Result
$2,130 a month on Claude Sonnet 5.5
$7.10 per 1,000 requests. Claude Haiku 5.5 would cost $108.00 for the same tokens.
- Cost per request
- $0.0071
- Cost per 1,000 requests
- $7.10
- Per month
- $2,130
- Per year
- $25,560
- Cheapest model for this workload
- Claude Haiku 5.5$108.00 a month
Worked examples
Each row is the calculator run on the inputs shown, with every other input at its default.
| Workload | Requests a day | Tokens in / out | Model | Per month |
|---|---|---|---|---|
| Support chatbot | 5,000 | 3,000 / 300 | Claude Sonnet 5.5 | $751.50 |
| Support chatbot on Haiku 5.5 | 5,000 | 3,000 / 300 | Claude Haiku 5.5 | $39.15 |
| Document summaries | 1,000 | 20,000 / 800 | Claude Sonnet 5.5 | $1,440 |
| Document summaries on Gemini 3.8 Flash | 1,000 | 20,000 / 800 | Gemini 3.8 Flash | $540.00 |
| Classification at scale | 200,000 | 400 / 5 | GPT-6 Luna | $147.00 |
| Coding agent | 2,000 | 60,000 / 2,000 | Claude Opus 5.5 | $4,488 |
| Coding agent on GPT-6.1 Sol | 2,000 | 60,000 / 2,000 | GPT-6.1 Sol | $2,244 |
How it is calculated
Each request is billed for its input tokens and its output tokens at the model's price per million tokens. Input that the provider serves from its prompt cache is billed at the lower cached price.
Prices are list prices for the standard tier of each provider's own API, checked on the date below. Batch and flex tiers usually cost half; regional or data-residency endpoints usually cost 10% more. Gemini 3.1 Pro and Claude Haiku 5.5 charge more for very long prompts; the calculator applies that when your input per request is over the threshold.
Two things move real bills away from this estimate. Thinking tokens are billed as output, so reasoning models can produce several times more output than the visible answer. And the same text becomes a different number of tokens on different models: Anthropic says Claude 4.7 and later produce about 30% more tokens than earlier Claude models for the same text.
Cache writes are not counted separately here. Anthropic and OpenAI charge more than the input price for the first write of a cached prompt; for a prompt reused thousands of times a day that cost is small.
Questions
How do I calculate the cost of an AI API call?
Multiply input tokens by the input price and output tokens by the output price, both per million tokens, and add them. Cached input is billed at a lower price. This page does that for every current model at once.
Which AI model is cheapest?
For most workloads the small models (GPT-6 Luna, Claude Haiku 5.5, Gemini 3.8 Flash) cost a few cents per thousand requests. The chart above sorts every model by cost for your exact workload.
How many tokens is a word?
In English, 1,000 tokens is about 750 words. Code, other languages and newer tokenizers produce more tokens per word.
Are Gemini prices going up?
Google lists higher prices for Gemini 4 Argon from December 1, 2026 and for Gemini 3.8 Flash from January 1, 2027. The calculator shows both the current and the future monthly cost when you pick one of those models.
Do thinking tokens cost extra?
They are billed as output tokens. Add them to the output tokens per request if your model reasons before it answers.
Sources
- Anthropic: Claude API pricing, checked 2026-10-29
- OpenAI: API pricing, checked 2026-10-29
- Google: Gemini Developer API pricing, checked 2026-10-29
Prices and plans change. This page was last checked on 2026-10-29. The data is open: llm-prices.json (CC BY 4.0).