OpenAI API pricing guide & calculator
OpenAI API prices range from $0.10 to $10 per million input tokens and $0.50 to $50 per million output tokens, depending on the model. Below: every current model’s price, what common jobs cost per month, and a calculator for your own usage.
Current OpenAI API prices
| Model | Input | Cached input | Output | Batch in / out |
|---|---|---|---|---|
| GPT-6 Astraflagship · 1.05M tokens | $10 | $1 | $50 | $5 / $25 |
| GPT-6.1 Solbalanced · 1.05M tokens | $2 | $0.10 | $10 | $1 / $5 |
| GPT-6 Lunafast · 1.05M tokens | $0.10 | $0.01 | $0.50 | $0.05 / $0.25 |
Per million tokens, standard tier, in US dollars. Checked October 10, 2026 against OpenAI’s official pricing. “Cached input” is the price for text you reuse across requests, like long instructions.
- GPT-6 Astra: OpenAI’s most capable model.
Estimate your monthly OpenAI API cost
Pick a job and how many you’ll run a month. OpenAI’s models are compared with Claude and Gemini; tap a provider to add or remove it. How the estimates work.
| Model | Per ticket | Per month | vs. cheapest |
|---|---|---|---|
| Claude Haiku 5.5Anthropic · fast | $0.00042 | $0.85 | 1× |
| GPT-6 LunaOpenAI · fast | $0.00042 | $0.85 | 1× |
| Gemini 3.5 Flash-LiteGoogle · fast | $0.0016 | $3.25 | 3.8× |
| Gemini 3.8 FlashGoogle · balanced | $0.0032 | $6.38 | 7.5× |
| Claude Sonnet 5.5Anthropic · balanced | $0.0085 | $17.00 | 20× |
| GPT-6.1 SolOpenAI · balanced | $0.0085 | $17.00 | 20× |
| Gemini 3.1 Pro (Preview)Google · flagship | $0.0092 | $18.40 | 22× |
| Claude Opus 5.5Anthropic · flagship | $0.0170 | $34.00 | 40× |
| Claude Fable 5.1Anthropic · flagship | $0.0425 | $85.00 | 100× |
| GPT-6 AstraOpenAI · flagship | $0.0425 | $85.00 | 100× |
Published list prices checked October 10, 2026, before tax. Token counts are estimates and vary by provider, since each counts text a little differently. Subscriptions like Claude Pro, ChatGPT Plus, or Google AI Pro (about $20 a month) are for using the chat apps yourself and can’t be called from your own software. Calculated in your browser; nothing is saved.
Which OpenAI model should you use?
OpenAI’s current lineup has three featured models, named after their place in the range. GPT-6.1 Sol is the everyday workhorse: OpenAI describes it as near-Astra performance at a lower cost, and at $2 per million input tokens and $10 per million output tokens it’s where most business apps should start.
GPT-6 Luna is the efficient option for focused, high-volume tasks, at $0.10 / $0.50. Classifying messages, extracting fields, short replies, and routing are all good Luna jobs. GPT-6 Astra is OpenAI’s most capable model, at $10 / $50, for the hardest reasoning and most demanding work.
OpenAI still sells many older models, including the GPT-5 family, at a wide range of prices. Some, like GPT-5 mini and nano, are cheap; others, like the “pro” models, cost far more than the current flagships. Unless you have a specific reason to stay on an older model, the GPT-6 family gives you better price for performance, and the calculator focuses on it.
What common jobs cost on OpenAI
Monthly cost at standard prices, with no caching or batch discounts, for typical volumes. The token sizes behind each job are explained in how we calculate.
| Job (per month) | GPT-6 Astra | GPT-6.1 Sol | GPT-6 Luna |
|---|---|---|---|
| 1,000 × draft an email | $28 | $5.6 | $0.28 |
| 2,000 × answer a support ticket | $85 | $17 | $0.85 |
| 300 × summarize a 10-page document | $31.5 | $6.3 | $0.315 |
| 100 × write a 1,500-word article | $11.2 | $2.24 | $0.112 |
| 5,000 × write a product description | $90 | $18 | $0.90 |
| 3,000 × chatbot conversation (10 turns) | $660 | $132 | $6.6 |
| 200 × ai coding or research agent task | $400 | $80 | $4 |
On GPT-6.1 Sol, drafting 1,000 emails a month costs about $5.6, and answering 2,000 support tickets about $17. The same jobs on GPT-6 Luna cost under a dollar each, which is why it’s worth testing Luna on any task that doesn’t need deep reasoning.
How OpenAI compares on price
OpenAI and Anthropic price their current lineups identically tier for tier, so the choice between GPT-6.1 Sol and Claude Sonnet 5.5 comes down to quality on your task, discounts, and long-prompt pricing rather than the list price. Google’s Gemini 3.8 Flash is cheaper in the middle of the range, and Gemini 3.1 Pro, Google’s flagship, is priced close to Sol rather than to Astra.
| Balanced model | Input / output per 1M | 2,000 tickets |
|---|---|---|
| Gemini 3.8 FlashGoogle | $0.75 / $3.75 | $6.38 |
| Claude Sonnet 5.5Anthropic | $2 / $10 | $17 |
| GPT-6.1 SolOpenAI | $2 / $10 | $17 |
Discounts: batch, caching, and free use
Batch and Flex: 50% off. OpenAI’s Batch API halves input and output prices for jobs that can wait, and the Flex tier offers the same prices for requests that can tolerate slower, lower-priority processing. Both suit overnight processing and bulk jobs.
Cached input: about 90–95% off. When the start of a prompt repeats across requests, OpenAI automatically charges the cached rate for that part: $0.10 instead of $2 per million on GPT-6.1 Sol, and $1 instead of $10 on Astra. You don’t have to set anything up, which makes it easy to benefit from; the trick is to put the parts that never change (instructions, examples, reference text) at the start of the prompt.
OpenAI also offers a Priority tier at higher prices for faster, more reliable processing, which the calculator doesn’t include.
Free tier: No free API tier.
What surprises people on the bill
Long prompts cost double
Requests with more than 272,000 input tokens are billed at OpenAI’s long-context rates: $4 / $15 instead of $2 / $10 on GPT-6.1 Sol, and $20 / $75 instead of $10 / $50 on Astra. The higher price applies to the whole request. Claude’s Sonnet, Opus, and Fable models don’t charge extra for long prompts, so for very long documents the comparison can swing.
Reasoning uses output tokens
Models that “think” before answering are billed for that thinking as output tokens, even though you don’t see all of it. On hard problems, output tokens, the expensive side of the price, can be many times longer than the visible answer. Measure real requests before you budget for reasoning-heavy work.
Old model names, new prices
OpenAI keeps older models available at their original prices, and some legacy models cost more than the current generation. If your app was built a year or two ago, check which model it calls; moving to a GPT-6 model can be both cheaper and better.
Regional processing costs extra
Data-residency (regional processing) endpoints add 10% for models released on or after March 5, 2026, and so do FedRAMP endpoints. The calculator uses standard global prices.
OpenAI API vs. ChatGPT Plus
ChatGPT subscriptions and the OpenAI API are billed separately. ChatGPT Plus ($20 a month) is for using ChatGPT yourself in the app. The API is for putting OpenAI’s models inside your own software, website, or automations, paid per token. A ChatGPT subscription doesn’t include API credits.
For comparison, $20 of API usage buys about 3,571 email drafts on GPT-6.1 Sol or roughly 71,428 on GPT-6 Luna, using our token estimates. If you’re writing for yourself, ChatGPT Plus is simpler. If you want AI to answer customers, process files, or run on a schedule without you, you need the API. Many people who outgrow ChatGPT’s chat app also compare it with Claude’s subscriptions, which include Claude’s desktop agent for longer tasks.
| OpenAI subscription | Price |
|---|---|
| ChatGPT Plus | $20/mo |
| ChatGPT Business | $25/mo (per seat; $20 billed annually) |
Who the OpenAI API is a good fit for
OpenAI’s API is the default for many developers because of its huge ecosystem: most tools, tutorials, and integrations support it first. Automatic prompt caching makes it easy to save money without extra setup, and the Flex tier is a simple way to halve costs for work that isn’t urgent.
Watch the long-prompt threshold if you work with large documents, and compare carefully with Claude, which matches OpenAI’s list prices tier for tier, and Gemini, which is cheaper in the middle of the range. Run your real prompts through two or three models before committing; the cheapest model that’s good enough wins.
Frequently asked questions
How much does the OpenAI API cost?
Per million tokens on the standard tier: GPT-6 Astra is $10 input / $50 output, GPT-6.1 Sol is $2 / $10, and GPT-6 Luna is $0.10 / $0.50, for prompts up to 272,000 tokens. Batch and Flex processing are half price, and cached input is much cheaper.
Does ChatGPT Plus include API access?
No. ChatGPT Plus is a subscription for using ChatGPT in the app. OpenAI API usage is billed separately, per token, through an OpenAI developer account.
What is the cheapest OpenAI model?
Among the current GPT-6 models, GPT-6 Luna at $0.10 per million input tokens and $0.50 per million output tokens. Some older models are cheaper still, such as GPT-5 nano, but are less capable.
How do I reduce my OpenAI API bill?
Use the smallest model that does the job well, put repeated instructions at the start of the prompt so they’re billed at the cached rate, send non-urgent work through Batch or Flex for 50% off, and keep prompts under 272,000 tokens to avoid long-context pricing.
Is OpenAI cheaper than Claude?
Their current list prices match tier for tier: GPT-6 Astra and Claude Fable 5.1 are both $10 / $50, GPT-6.1 Sol and Claude Sonnet 5.5 are both $2 / $10, and GPT-6 Luna and Claude Haiku 5.5 are both $0.10 / $0.50. Real costs differ with long prompts, caching, and how many tokens each uses for the same text.