OpenAI API and ChatGPT cost calculator: token-based costs
How much does using GPT models through OpenAI API cost? API usage costs are determined primarily by input and output token counts, the selected model and additional options such as Prompt Caching, Batch, Flex, Fast and tools such as Web Search.
Choose a model, enter your estimated input and output token counts and the number of requests, and the calculator will show an estimated cost in dollars.
It is important to distinguish between OpenAI API and a ChatGPT subscription. A ChatGPT Plus, Pro, Business or Enterprise subscription does not automatically become an API balance, and API platform billing is handled separately.
OpenAI uses the term credits in several contexts. For OpenAI API, credits usually represent a monetary balance from which usage costs are deducted. There is no fixed ratio equating one credit to a particular number of tokens.
OpenAI API and ChatGPT cost calculator: token-based costs
Prices last checked: September 23, 2026
Long-context pricing applied For certain OpenAI models, a request with more than 272,000 input tokens is billed at long-context rates for the entire request.
- Estimated cost per request
- Estimated total cost
- Input token cost
- Output token cost
- Cache read
- Cache write
- Web Search
- Estimated cost for 100 identical requests
- Estimated cost for 1,000 identical requests
- Processing type
- Pricing
This calculation is an estimate only, based on OpenAI list prices and the data entered. Actual charges are determined by usage recorded by OpenAI and may vary due to caching, reasoning tokens, long context, tools, processing type and pricing changes.
How is OpenAI API pricing calculated?
For most OpenAI text models, costs depend on the number of input tokens the model receives and output tokens it generates. Input tokens may include instructions, conversation history, code, documents, tool definitions and schemas, and other information sent in the request.
Output tokens include the text the model returns and reasoning tokens billed as output when used. Calculator prices are shown in dollars per million tokens.
What is the difference between input and output tokens?
Input tokens are the information sent to the model. Output tokens are the information the model generates.
Each model may have different input and output pricing, and in some models output tokens are significantly more expensive than input tokens. Two requests with the same input token count can therefore cost different amounts if one produces a much longer response.
What are OpenAI API credits?
OpenAI allows some API accounts to purchase a prepaid usage balance. These credits are not a token unit.
There is no fixed ratio equating one credit to a particular number of tokens, because models have different prices and token type also affects pricing. The calculator therefore uses dollars as its main cost unit.
Do ChatGPT Plus or Pro subscriptions include API usage?
Not automatically. ChatGPT and the OpenAI API platform use separate billing systems.
A ChatGPT subscription may grant access to models and capabilities within ChatGPT, but it does not necessarily create an API balance usable through an API key. Anyone using OpenAI API directly needs to configure API platform billing according to their account terms.
What is Prompt Caching?
Prompt Caching lets OpenAI reuse matching parts of input that have already been processed. When a cache hit occurs, certain tokens may be billed at a reduced cached input rate, below the standard input rate.
GPT-5.6 and GPT-6 models also have a separate cache-write rate. A cache write is not a surcharge added to the standard input rate for the same token. Each token is assigned to its applicable billing category: standard input, cache read or cache write.
The calculator lets you enter this breakdown under advanced options.
Why are cached tokens not added to total input?
In OpenAI usage data, tokens classified as cached are part of the total input token count. If a request includes 20,000 input tokens, of which 15,000 came from the cache, you should not calculate 35,000 tokens.
The calculator subtracts cache-read or cache-write tokens from total input to calculate the amount of standard input.
How are long-context requests priced?
Some models have a special pricing tier for very large requests. In supported calculator models, when input exceeds 272,000 tokens, long-context pricing may apply to the entire request rather than just tokens above the threshold.
The calculator detects the threshold automatically for models where OpenAI has published such a rule. Not every model with a large context window uses this pricing.
What is the difference between Standard, Batch, Flex and Fast?
Standard is the normal processing tier. Batch is designed for asynchronous processing and can reduce token costs, but is not intended for requests requiring an immediate response.
Flex may offer lower token pricing in exchange for less predictable response times or availability, and only on supported models. Fast is designed for faster processing but uses higher token rates.
The calculator displays only options verified for the selected model.
Do reasoning tokens cost more?
There is no separate reasoning token rate simply because they are thinking tokens. Reasoning tokens are billed as output tokens at the model's output rate.
However, a model using more reasoning tokens can generate more billable output tokens even when the final answer shown to the user is relatively short. An estimate based solely on the visible answer's length can therefore be lower than actual charges.
How much does Web Search cost?
The checked Web Search rate is 10 dollars per 1,000 successful calls, or 0.01 dollars per successful call. This is not necessarily the full cost of using search, however.
Content returned by search and sent to the model can add input tokens billed at the model's rate.
Why can actual costs differ?
The calculator estimates costs based on the data entered and checked list prices. Actual charges may differ due to:
- The actual token count measured by the API
- Conversation history sent again
- Reasoning tokens
- Caching
- Tools
- Web search
- Files and images
- Long context
- Processing type
- Pricing changes
The calculator excludes taxes, currency conversions, custom Enterprise pricing and third-party provider prices.
What about Azure OpenAI?
The calculator covers direct OpenAI API usage only. Azure OpenAI is priced through Microsoft, and prices may vary by region, deployment type and commercial agreement.
Do not use this calculator's result as a price estimate for Azure OpenAI or other third-party providers.
When were prices checked?
The calculator's prices and pricing rules were checked against official OpenAI sources on September 23, 2026.
GPT-6 Sol and GPT-6 Luna launched on September 22, 2026. At the time prices were checked, official model pages had been published, but some of OpenAI's general summary pages had not yet been fully updated.
Model prices and terms of use can change. Before large-scale commercial use, check current pricing on OpenAI's website as well.