OpenAI API Pricing (2026): GPT Token Rates & API Costs

OpenAI API pricing is usage-based: you pay per million tokens, with separate rates for input, cached input and output, and the rate depends on the model and the processing speed you choose. As of November 10, 2026, OpenAI's flagship GPT-6 Astra costs $10 per million input tokens and $50 per million output tokens, the mid-priced GPT-6.1 Sol costs $2 / $10, and the most efficient model, GPT-6 Luna, costs $0.10 / $0.50.

This guide is the OpenAI API price sheet in plain English: the official token rates, what Batch, Flex, Fast and Ultrafast do to the price, how long context changes it, the tool fees, and a method to estimate GPT API costs for your own app. All figures come from OpenAI's official API pricing page and models page, checked on November 10, 2026, in US dollars. Rates change, so confirm before you budget.

A ChatGPT subscription is a different product with different billing — more on that below.

What OpenAI API pricing covers (vs a ChatGPT plan)

The OpenAI API is for developers calling OpenAI models from their own code, through the Responses API and OpenAI's SDKs. You are billed on tokens and tool usage.

ChatGPT plans — Free, Go, Plus ($20/month) and Pro ($100, $200 or $500/month) — are subscriptions to the ChatGPT app. OpenAI's Plus help article is explicit: "API usage is separate and billed independently." OpenAI's docs do have a page on API credits for ChatGPT subscribers, so check your eligibility there, but a ChatGPT plan is not an API plan. For plan comparisons, see Claude vs ChatGPT and Gemini vs ChatGPT.

OpenAI API price list: official rates at a glance

These are the four featured models on OpenAI's models page, with standard rates for short context (up to 272K input tokens).

OpenAI's own starting advice: use GPT-6 Astra for complex reasoning and coding if you are not sure, GPT-6.1 Sol to balance intelligence and cost, or GPT-6 Luna for cost-sensitive, high-volume workloads.

Other current and older models

The same sheet lists more models. A selection (standard, short context):

Model	Input	Output
gpt-6-sol	$2.00	$10.00
gpt-5.6-sol	$4.00	$20.00
gpt-5.6-terra	$2.00	$12.00
gpt-5.6-luna	$0.20	$1.20
gpt-5.5	$5.00	$30.00
gpt-5.4-mini	$0.75	$4.50
gpt-5-nano	$0.05	$0.40
gpt-4.1-mini	$0.40	$1.60

OpenAI notes that GPT-5.6 Sol's current price is promotional until November 30, 2026. Check the live sheet if you plan to use it after that.

How GPT API costs add up

Your bill is the sum of several parts. Each one is on the official sheet.

1. Input, cached input and output tokens

Output is five times input on all four featured models. Cached input — repeated prompt prefixes OpenAI has cached — is billed at a tenth of input or less (for example $0.10 vs $2.00 on gpt-6.1-sol). The sheet also lists cache write prices for 30-minute and 12-hour caches (for gpt-6.1-sol, $2.50 and $5.00 per million tokens).

2. Long context

Requests with more than 272K input tokens use long-context rates. For example, gpt-6.1-sol rises from $2 / $10 to $4 / $15, and gpt-6-astra from $10 / $50 to $20 / $75.

3. Processing speed: Batch, Flex, Fast and Ultrafast

Batch and Flex are half the standard price. Fast is double. Ultrafast is listed only for gpt-6-astra and gpt-6.1-sol. OpenAI notes that "Priority processing" was renamed Fast mode on July 30, 2026.

4. Data residency

OpenAI charges a 10% uplift for regional processing (data residency) endpoints on models released on or after March 5, 2026, and the same uplift on FedRAMP endpoints.

5. Tools

Tool	Official price
Web search	$10.00 per 1,000 calls, plus search content tokens at model rates
File search	$0.10 per GB per day storage (1 GB free); $2.50 per 1,000 tool calls
Hosted Shell / Code Interpreter containers	From $0.03 (1 GB) to $1.92 (64 GB) per 20-minute session

How to estimate a GPT API bill

OpenAI does not publish an "average" API bill, and a guessed one would mislead you. Use this method:

1. Measure tokens for a typical request with OpenAI's token counting tools, on the model you plan to use.
2. Split input, cached input and output. If your system prompt repeats on every call, much of it may be billed as cached input after the first request.
3. Apply the rates for your model, context length and processing tier.
4. Multiply by volume per month.
5. Add tool calls such as web search, plus the 10% uplift if you use a data-residency endpoint.

Worked example (arithmetic only, not a measured workload): a request with 3,000 input tokens and 800 output tokens on gpt-6.1-sol at standard rates costs (3,000 ÷ 1,000,000 × $2) + (800 ÷ 1,000,000 × $10) = $0.006 + $0.008 = $0.014. At 1,000 requests a day for 30 days, that is about $420 a month. The same traffic on gpt-6-luna costs $0.0003 + $0.0004 = $0.0007 per request, or about $21 a month. Run through the Batch API, both figures halve.

To test a prompt without code, try our AI token calculator.

Ways to cut OpenAI API costs

These are the levers on OpenAI's own sheet. We do not promise a percentage saving, because it depends on your traffic.

• Use Batch or Flex for work that can wait: half the standard price.
• Keep prompts stable at the start so repeated prefixes are billed as cached input.
• Stay under 272K input tokens where you can, to avoid long-context rates.
• Pick the smallest model that passes your tests. gpt-6-luna is a hundredth of gpt-6-astra's input price.
• Cap output length, since output is the expensive side.
• Use Fast or Ultrafast only where latency matters — they cost 2x and up.

OpenAI API vs Claude API vs Gemini API

These are not like-for-like quality comparisons, only the official prices of each provider's mid-range option on the same day. Usage mix, caching and output length change the real bill. For the other sheets, see Claude API pricing and Gemini API pricing. For low-cost options from Chinese providers, see Chinese AI models, and for assistants beyond OpenAI, our ChatGPT alternatives guide.

FAQ