OpenAI API cost in 2026, next to Claude and Gemini
Current per-token prices for OpenAI, Anthropic and Google models, taken from their official pricing pages on November 11, 2026, with worked examples of what one user costs you.
How much does the OpenAI API cost?
The OpenAI API charges per million tokens, and the price depends on the model. As of November 2026, GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens, GPT-6.1 Sol costs $2 and $10, and the flagship GPT-6 Astra costs $10 and $50. Batch requests cost half.
- Cheapest current OpenAI model
- GPT-6 Luna: $0.10 in, $0.50 out per 1M tokens
- OpenAI flagship
- GPT-6 Astra: $10 in, $50 out per 1M tokens
- Cheapest Claude model
- Claude Haiku 5.5: $0.10 in, $0.50 out per 1M tokens
- Gemini 3.8 Flash
- $0.75 in, $3.75 out per 1M tokens until December 31, 2026
- Batch discount
- 50% at OpenAI, Anthropic and Google
- One token
- About 4 characters of English text
How much do the OpenAI, Claude and Gemini APIs cost?
AI APIs charge per million tokens, with separate prices for what you send (input) and what the model writes back (output). This table lists current text models from the official pricing pages of OpenAI, Anthropic and Google, checked on November 11, 2026. Prices are standard, pay-as-you-go rates in US dollars.
| Provider | Model | Input | Output | Cached input |
|---|---|---|---|---|
| OpenAI | GPT-6 Astra | $10.00 | $50.00 | $1.00 |
| OpenAI | GPT-6.1 Sol | $2.00 | $10.00 | $0.10 |
| OpenAI | GPT-6 Sol | $2.00 | $10.00 | $0.20 |
| OpenAI | GPT-6.1 Terra | $1.00 | $5.00 | $0.05 |
| OpenAI | GPT-5.6 Terra | $2.00 | $12.00 | $0.20 |
| OpenAI | GPT-5.6 Luna | $0.20 | $1.20 | $0.02 |
| OpenAI | GPT-6 Luna | $0.10 | $0.50 | $0.01 |
| Anthropic | Claude Fable 5.2 | $8.00 | $40.00 | $0.20 |
| Anthropic | Claude Fable 5.1 | $10.00 | $50.00 | $0.25 |
| Anthropic | Claude Opus 5.5 | $4.00 | $20.00 | $0.20 |
| Anthropic | Claude Sonnet 5.5 | $2.00 | $10.00 | $0.10 |
| Anthropic | Claude Haiku 4.5 | $1.00 | $5.00 | $0.10 |
| Anthropic | Claude Haiku 5.5 | $0.10 | $0.50 | $0.01 |
| Gemini 4 Argon | $2.00 | $10.00 | $0.10 | |
| Gemini 3.8 Flash | $0.75 | $3.75 | $0.075 | |
| Gemini 3.5 Flash-Lite | $0.30 | $2.50 | $0.03 | |
| Gemini 3.1 Flash-Lite | $0.25 | $1.50 | $0.025 |
- OpenAI prices are for prompts up to 272,000 tokens; longer prompts cost more, for example $20 in and $75 out on GPT-6 Astra.
- Claude Haiku 5.5 prices are for prompts up to 100,000 tokens; above that it costs $0.50 in and $2.50 out.
- Gemini 4 Argon’s prices double on December 1, 2026, to $4 in and $20 out. Gemini 3.8 Flash’s double on January 1, 2027, to $1.50 and $7.50.
- Google counts thinking tokens as output. Gemini 4 Argon is not on Google’s free tier; several other Gemini models have one, where Google may use your content to improve its products.
- Caching also has write prices, which this table leaves out: OpenAI charges 1.25 times the input rate for a 30-minute cache write on its newer models, and Anthropic $5 per million for a 5-minute write on Claude Opus 5.5.
What is a token, and why does output cost more?
A token is a chunk of text the model reads or writes. Google’s token guide puts it at about 4 characters, or 60 to 80 English words per 100 tokens. Short common words are often one token; long or rare words split into several.
You pay for every token in both directions. Input includes your instructions, the conversation so far and any documents you attach, so it grows with each turn of a chat. Output is what the model writes, and on most models in the table it costs five times the input price.
How much does AI cost per user?
For a typical chat feature, from a fraction of a cent to a few dollars per active user a month, depending on the model. The worked examples below are our own illustrations, not vendor figures. They assume each message sends 1,500 input tokens (instructions, history and the question) and gets 300 output tokens back, and that an active user sends 10 messages a day for 30 days: 300 messages a month.
| Model | Per message | Per user a month | Per 1,000 users a month |
|---|---|---|---|
| GPT-6 Luna | $0.0003 | $0.09 | $90 |
| Claude Haiku 5.5 | $0.0003 | $0.09 | $90 |
| Gemini 3.8 Flash | $0.00225 | About $0.68 | About $675 |
| GPT-6.1 Sol | $0.006 | $1.80 | $1,800 |
| Claude Sonnet 5.5 | $0.006 | $1.80 | $1,800 |
| GPT-6 Astra | $0.03 | $9.00 | $9,000 |
How we worked it out, for GPT-6.1 Sol: 1,500 input tokens at $2 per million is $0.003, and 300 output tokens at $10 per million is $0.003, so $0.006 a message. Times 300 messages is $1.80 a user a month. Your numbers will differ with longer chats, documents or images, so measure real token counts once you have users. For the full build-and-run budget, see what it costs to make an app.
The model choice is the biggest lever: the same feature costs 100 times more on GPT-6 Astra than on GPT-6 Luna. That is why pricing an AI product on flat subscriptions alone is risky; see usage-based pricing for how products pass these costs on.
How can you cut AI API costs?
Use the cheapest model that does the job, then use the discounts each vendor documents. These four come from the vendors’ own docs:
- Smaller models: GPT-6 Luna, Claude Haiku 5.5 and the Gemini Flash-Lite models cost a small fraction of the flagships. Test them first and only move up when quality falls short.
- Prompt caching: reused parts of a prompt, like a long system prompt, are billed at the cached-input price. OpenAI caches automatically on supported models, from 1,024 tokens on GPT-5.6 and later; Anthropic bills cache hits at 0.1 times the input price, 0.05 times on Claude Opus 5.5 and Sonnet 5.5, and 0.025 times on Claude Fable 5.1 and 5.2; Google applies context caching too.
- Batch processing: jobs that can wait cost 50% less at all three. OpenAI’s Batch API completes within 24 hours.
- Flex processing: OpenAI and Google offer slower, lower-priority processing at batch rates on supported models.
| Scenario | Standard cost | With the discount |
|---|---|---|
| Chat message with a 1,200-token cached system prompt (300 fresh input, 300 output tokens) | $0.006 | About $0.0037 with caching, before cache writes |
| 10,000 product descriptions, 500 input and 400 output tokens each | $50 | $25 through the Batch API |
Watch for costs that go the other way: OpenAI adds 10% for regional data residency endpoints on newer models, Anthropic applies a 1.1 times multiplier for US-only inference, and OpenAI’s Fast mode and Google’s Priority tier cost more than standard.
How much do AI chatbot tools charge compared with the raw API?
Ready-made AI chatbot tools charge far more per conversation than the raw API, because you are paying for the product around the model: training on your content, a helpdesk, integrations and analytics. Prices from their own pages, as of November 2026:
| Option | What you pay | Billed |
|---|---|---|
| Raw API, GPT-6 Luna (our example message) | About $0.0003 per message | Per token, no minimum listed |
| Raw API, GPT-6.1 Sol (our example message) | About $0.006 per message | Per token, no minimum listed |
| Chatbase Hobby | $40 a month ($32 billed yearly) for 700 message credits | Extra credits $40 per 1,000 |
| Intercom Fin | $0.99 per outcome | Minimum 50 outcomes a month with your own helpdesk |
Sources: Chatbase pricing and Fin pricing. The raw API is cheapest by far, but you build and run everything else yourself. A middle path is an AI wrapper: your own product on top of the API, priced so each user pays more than their AI cost.
Does an OpenAI API key cost money?
OpenAI lists no fee for the API key itself; you pay for the tokens you use with it. You pay the input and output rates of the model you call. Built-in tools such as web search and file search have their own charges on the same page.
If you’d rather not manage keys, bills and rate limits at all, an AI app builder can run the models for you and meter the cost inside your plan. Our guide to the cost to make an app covers the rest of the bill.
Starty: AI features without API keys or surprise bills
Starty runs the AI in your app for you, from a ready-made catalog of capabilities, and shows what each user costs next to what they pay. It is credit-based and free to start.
- Switch on AI capabilities like agents, vision, voice, image and video in one click, with no API keys of your own
- Every use is metered against your credits, so there is no separate AI bill
- Each user’s plan, AI usage and AI cost in one view, so you know what each customer costs you
- Free to start, with credit-based plans and no Starty fee on your revenue
Related guides
Where every fact comes from
Official pricing pages, docs, announcements and named studies, checked on November 11, 2026. Where a company doesn’t list something, we say so rather than guess.
Questions people ask
There is no monthly fee listed; you pay per token. As of November 2026, GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens, and GPT-6.1 Sol costs $2 and $10. Your bill depends on how many tokens your app sends and receives.
OpenAI lists no charge for the key itself. You pay for the tokens your requests use, at the rates on its pricing page, plus any built-in tools you call.
Among each vendor’s current lineup on the official pricing pages in November 2026, GPT-6 Luna and Claude Haiku 5.5 tie at $0.10 per million input tokens and $0.50 per million output tokens. Batch processing halves those prices.
In our example of 300 short messages a month, one active user costs about $0.09 on GPT-6 Luna and $1.80 on GPT-6.1 Sol in raw API fees. Chatbot tools charge more, for example Intercom Fin at $0.99 per outcome.
Roughly 600,000 to 800,000 English words, using Google’s rule of thumb that 100 tokens is about 60 to 80 words. Other languages and other vendors’ tokenizers give different counts.
It depends on the model. Claude Sonnet 5.5 and GPT-6.1 Sol both cost $2 in and $10 out per million tokens, and Claude Haiku 5.5 matches GPT-6 Luna. Claude Opus 5.5 at $4 and $20 sits below GPT-6 Astra at $10 and $50.
Enter your email to join Starty. We’ll email you your invite, and you can start building for free.
No. You describe what you want, or fork an idea, and Starty builds it. You review and approve; you never have to open the code.
No. Starty takes no fee on your revenue. Payouts go to your bank account, and Starty earns from plans and credits.