Estimate your monthly cost
Prices last checked: …. Standard tier, USD per 1M tokens.
Presets:
| Model | Input / output per 1M | Per request | Per month (30 days) |
|---|
Official sources:
How the calculator works
AI APIs charge by the token, with separate prices for input (what you send: instructions, conversation history and documents) and output (what the model writes back). Prices are quoted per million tokens. The calculator multiplies your token counts by each model’s published rates, then by your daily volume and 30 days:
monthly cost = (input tokens × input price + output tokens × output price) ÷ 1,000,000 × requests per day × 30
If you set a cached-input percentage, that share of the input is charged at the model’s cached rate (where the provider lists one). The table sorts models from cheapest to most expensive for your exact workload, because the cheapest model for short questions isn’t always the cheapest for long documents.
Where the prices come from
All prices live in a single data file that we update by hand from the official pricing pages linked above, and the “last checked” date is shown at the top of the tool. We list each provider’s standard, pay-as-you-go rate for prompts within the normal context length. Some prices are temporary: Google’s pricing page lists Gemini 4 Carbon at $2.00 input / $10.00 output through November 30, 2026, rising to $4.00 / $20.00 from December 1, 2026, and Gemini 3.8 Flash rising from January 1, 2027. DeepSeek charges different rates at peak and off-peak hours; we show the higher peak rate so estimates err on the safe side. Claude Haiku 5.5 is shown at its rate for prompts up to 100,000 tokens; Anthropic charges more above that.
How to estimate your token numbers
- Measure a real prompt. Paste a typical request — including the system instructions and any document you attach — into the token counter.
- Estimate the answer length. A 300-word answer is roughly 400 tokens by OpenAI’s rule of thumb (0.75 words per token).
- Remember conversation history. In a chat app, each new message usually re-sends earlier turns as input. A 10-turn conversation can cost far more than ten separate questions.
- Add a safety margin. Real usage is spiky. Budget 20–30% above your estimate until you have real billing data.
Worked example: a small support chatbot
Imagine a shop’s help-desk bot that receives 200 questions a day. Each request sends about 500 input tokens (instructions plus the question) and gets a 300-token answer. Click the “Support Chatbot” preset to load these numbers. With the prices in our data file, a low-cost model like GPT-6 Luna or Claude Haiku 5.5 costs only a few dollars a month for this workload, while a flagship model costs several times more. The question is whether the flagship’s better answers are worth the difference for your customers — which is something only testing on your real questions can tell you.
Remove this box by adding your content here, or hide all notes with
"showEditorNotes": false in site.config.json.Ways to cut API costs
- Use the smallest model that does the job. Route simple tasks (classification, short replies) to a small model and reserve flagship models for hard reasoning.
- Trim the prompt. Remove repeated instructions and irrelevant context. Every token you send is billed.
- Cap the output. Ask for concise answers and set a maximum output length in your API call.
- Use caching and batch modes. OpenAI, Anthropic and Google all describe batch processing at about half the standard price for work that doesn’t need an instant answer, plus discounted cached input.
- Summarise long conversations instead of re-sending the full history every turn.
Limitations
This is a planning tool, not a quote. It doesn’t include taxes, minimum charges, tool-use fees (such as web search or code execution), image or audio pricing, long-context surcharges, or enterprise discounts. Token counts are estimates. Always confirm against the provider’s pricing page and your real billing dashboard before making business decisions.
Frequently asked questions
Are these API prices up to date?
They were copied from each provider’s official pricing page on the date shown above the table. AI prices change often — Google, for example, has announced increases for some Gemini models in December 2026 and January 2027 — so check the linked source before you commit to a budget.
Why is output more expensive than input?
Providers price generated tokens higher than tokens you send, because generating text takes more computation. For most models in our list, output costs four to five times as much as input per million tokens.
Does this include taxes, batch discounts or free tiers?
No. It uses standard, pay-as-you-go rates for short prompts. Batch processing (often about 50% cheaper), long-context surcharges, free tiers, regional pricing and taxes are not included.
What is cached input?
Some providers charge less when the start of your prompt repeats across requests and is served from a cache. Set the cached percentage to estimate the saving. If a model’s cache price isn’t listed in our data, the full input price is used.
Is a chatbot subscription the same as API pricing?
No. Consumer plans like ChatGPT Plus or Claude Pro are monthly subscriptions with usage limits. API pricing is pay-per-token for developers building apps. See our free vs paid plans guide for subscriptions.
Sources
- OpenAI API pricing (opens in a new tab)
- Anthropic Claude API pricing (opens in a new tab)
- Google Gemini API pricing (opens in a new tab)
- DeepSeek API pricing (opens in a new tab)
Facts, prices and features were checked against these sources on November 19, 2026. AI products change quickly; if you spot something out of date, please tell us.