OpenAI API Pricing Explained (2026): Per Million Tokens, on a Different Bill from ChatGPT

The OpenAI API bills tokens, tool calls, and a few session meters. It does not bill a ChatGPT seat. On the markdown form of the pricing page, fetched October 24, 2026 at developers.openai.com/api/docs/pricing.md, standard short-context prices per 1 million tokens include gpt-6-astra at $10 input and $50 output, gpt-6.1-sol at $2 and $10, and gpt-6-luna at $0.10 and $0.50. Cached input on those three rows is $1, $0.10, and $0.01. Short context on that page means 272,000 input tokens or fewer. Above that, a second set of columns applies.

A ChatGPT subscription is the other product. Plus is $20 a month on learn.chatgpt.com/docs/pricing the same day. The seat ladder is ChatGPT pricing explained. Image rows on the API page are walked in DALL·E pricing explained. The idea of a key in front of a product is what an API for AI tools is.

Published October 24, 2026. We did not run a bill. Prices below are the ones on that markdown page. The directory has no separate OpenAI API card, so the product link for the chat side of the company is ChatGPT.

How to read one row

What you pay	gpt-6-astra, standard, short context	gpt-6.1-sol, standard, short context	gpt-6-luna, standard, short context
Input, per 1M tokens	$10	$2	$0.10
Output, per 1M tokens	$50	$10	$0.50
Cached input, per 1M tokens	$1	$0.10	$0.01
Cache write, 30 minutes / 12 hours, per 1M	$12.50 / $25	$2.50 / $5	$0.125 / $0.25

flowchart TD
  job{"What are you paying for?"}
  job -->|A person in ChatGPT| seat["A ChatGPT plan. Plus is 20 USD a month"]
  job -->|Your app, standard speed, short context| std["The standard per-1M-token row for the model you call"]
  job -->|The same model, batch or flex| half["A lower column. Astra short-context batch is 5 USD input and 25 USD output"]
  job -->|A tool on top of the tokens| tool["The tool price plus the model tokens"]

Long context, above 272,000 input tokens, is a different price. On the same standard table, gpt-6-astra long-context input is $20 per 1M tokens and long-context output is $75. Do not apply the $10 input price to a prompt that has crossed the line. The page also says regional processing endpoints charge a 10 percent uplift for models released on or after March 5, 2026, and that FedRAMP endpoints charge a 10 percent uplift. Priority processing was renamed Fast mode on July 30, 2026, on that page.

Batch and Flex, for gpt-6-astra short context, are $5 input, $25 output, and $0.50 cached input on that page. Fast for the same cell is $20 input, $100 output, and $2 cached input. Ultrafast is a further column. Quote the column you will actually call. A batch discount you hoped for is what AI batch inference is only when you are on the batch column.

Tools and seats that are not the token row

Web search, on that page, is $10 per 1,000 calls for the all-models line, plus search content tokens at model rates. The non-reasoning web-search preview line is $25 per 1,000 calls, and the page says search content tokens are free on that preview line. Read the line your request uses.

File search is $0.10 per GB per day after 1 GB free, and $2.50 per 1,000 tool calls, on that page. The page says the tool-call price applies to the Responses API.

Containers for Hosted Shell and Code Interpreter are listed as $0.03 for 1 GB, $0.12 for 4 GB, $0.48 for 16 GB, and $1.92 for 64 GB, per 20-minute session per container. The same page also says eligible container sessions are billed by the minute with a 5-minute minimum. Use both sentences when you forecast a session.

A ChatGPT Business seat is $20 a user a month billed annually, for two or more users, or $25 monthly. It is a workspace. It is not a prepaid token balance. The workspace article is ChatGPT Team pricing explained.

A neighboring API, so the shape of a token bill is not unique to OpenAI: DeepSeek Flash off-peak is $0.15 per million input tokens on a cache miss on the directory card. That is DeepSeek’s meter. The invoice is DeepSeek pricing explained.

A coding seat that can call models without you holding this key is GitHub Copilot Pro at $10 a user a month on the October 16 card. Copilot credits are not this API price list. The company seat is GitHub Copilot Business pricing explained.

What not to assume

Do not add a Plus seat and an API forecast and call the sum “ChatGPT.” Do not price a call with only the input column. Output, cached input, and cache writes are on the same row. Do not treat a cached-input price as the price of the first call. The first call is the miss, until you have measured a hit. Do not copy an image price into a chat forecast. Image models are a separate table on the same page.

FAQ

How much is the OpenAI API?

There is no single price. On October 24, 2026, standard short-context gpt-6-luna was $0.10 per 1M input tokens and $0.50 per 1M output tokens. gpt-6.1-sol was $2 and $10. gpt-6-astra was $10 and $50. Your model and your speed column decide the row.

Does ChatGPT Plus include the API?

Plus is $20 a month for the ChatGPT product. API usage bills on the API account. A person who has Plus and an API key has two bills. The Codex plan page says an API key pays API pricing.

What is short context?

The pricing page says short context is 272,000 input tokens or fewer, and long context is above that. Astra’s long-context standard input on that page is $20 per 1M tokens, not $10.

Are Batch and Standard the same model at a discount?

They are different columns. Astra’s short-context batch input is $5 per 1M tokens on the page we read, against $10 on standard. You get the batch price when the request is actually batch or flex, as that table defines it. A normal call does not inherit the batch cell.

Last verified

October 24, 2026, from https://developers.openai.com/api/docs/pricing.md, the markdown form of the OpenAI API pricing page. ChatGPT seat prices in the comparison are from learn.chatgpt.com/docs/pricing that day. DeepSeek and GitHub Copilot comparisons use the October 16 directory cards. Published October 24, 2026.

Sources: OpenAI API pricing, ChatGPT and Codex pricing, DeepSeek API pricing, GitHub Copilot.

Published by Andrew for AI Tools Optimizer.

ChatGPT is in the chatbots and assistants directory. Directory pages may use affiliate links. A purchase through them may earn us a commission at no extra cost to you.