Grok API Pricing and Billing (November 2026): Credits, Invoices and Subscriptions
APIsQuick answer: The Grok API is usage-based. Text models bill per million tokens; image and video models bill per image or per second of video. A SuperGrok plan or The X Subscription can include API access, but x.ai/pricing says usage is billed separately at standard model rates and does not come out of the plan’s usage pool. You pay from prepaid credits in the xAI console, or from monthly invoiced billing if you enable it. Live rates are on xAI’s API pricing page (checked November 18, 2026).
xAI’s developer docs now carry the SpaceXSI name; the console and API domains are still x.ai.
How is the Grok API priced?
From the API pricing and models pages:
| Kind of request | How it is billed | Official page |
|---|---|---|
| Text models (Grok 4.7 and other listed chat/code models) | Per million input, cached input and output tokens | API pricing |
| Long-context text requests | Once the prompt reaches the listed threshold (200k tokens on the models that use this), the higher long-context rate applies to all tokens in that request | API pricing |
| Image generation | Per image, varying by model and quality | API pricing |
| Video generation | Per second of video, varying by model and resolution | API pricing |
| Voice | Separate speech-to-speech, speech-to-text and text-to-speech rates | API pricing |
| Server-side tools (web search, X search, code execution and others) | Token usage plus, for some tools, a per-call or per-item charge | API pricing |
xAI recommends Grok 4.7 for code, chat and everything except dedicated media APIs. Use the Imagine image and video models for pictures and video, and the Voice API for speech. The knowledge cutoff for Grok 4.7, per the models page, is May 2026; for current events you need the search tools, which add to the bill.
We do not reprint per-million-token rates. They differ by model and can change; use the live table.
Does a Grok subscription pay for API usage?
No. This is the usual point of confusion with queries like “grok api billing”.
x.ai/pricing says paid plans include Grok API access, but API usage is billed separately at standard model rates and does not draw from your plan’s usage pool. The weekly pool on The X Subscription covers eligible first-party Grok and Cursor product usage, not console API traffic.
So you may need both: a consumer or team plan if you use Grok or Cursor as a product, and console credits if you call https://api.x.ai/v1. Creating a key is covered in How to get a Grok API key. For what SuperGrok and the legacy tiers still mean, see What is SuperGrok?.
How do prepaid credits and invoiced billing work?
The console billing docs and the billing FAQ describe two options:
- Prepaid credits. You buy credits for a team; each API request is priced immediately and deducted from that balance. This is the usual way to use the API and to cap spend. Auto top-up can buy more credits when the balance falls below a threshold you set. Bank transfers take a few business days before credits appear.
- Monthly invoiced billing. Off by default. If you leave the invoiced limit at zero, the API uses only prepaid credits and rejects requests when credits run out. If you raise the limit, prepaid credits are still used first, and the rest is charged to your default payment method at month-end, up to that limit.
The billing FAQ also notes:
- Indian payment cards cannot be saved for later charges, because of regulatory rules. Buy prepaid credits through Guest Checkout instead.
- Prepaid credits are not refundable, except where the law requires it.
- xAI cannot retroactively regenerate an invoice after you change billing details, so set the address and tax information correctly first.
Invoices and payment methods live under Billing in the xAI console. Make sure you are in the right team before you change anything: billing changes apply to everyone on that team.
What extra rates should you know about?
These rules are on the API pricing page. They change the bill without being obvious from the base token table:
- Batch API. Asynchronous jobs, usually within 24 hours, at a discount on listed text models. The discount applies to all token types for those models. Image and video can run through Batch but are billed at standard rates. Batch requests do not count against per-minute rate limits.
- Priority processing. A premium over standard token rates for lower-latency text requests on Chat Completions and Responses. You are charged the priority rate only when the response reports that the request was served on the priority tier.
- US regional endpoint. Requests to
https://us.api.x.ai/v1run in the United States and are billed at a premium to global rates. Only some models are available on that endpoint. The global endpoint remainshttps://api.x.ai/v1. - Grok 4.7 Fast is not on the public API. It is a faster serving of Grok 4.7 in Cursor and Grok Build, billed through the plan there.
- Files and collections stored on the platform have storage and download charges, listed on the same pricing page.
- If a request is blocked as a usage-guideline violation, xAI may still charge for that request, including a per-request fee in some Responses API cases. See the pricing page for the current rule.
How can you keep Grok API costs down?
- Prefer cached input where the API supports it; cached tokens are cheaper than fresh input on the official table.
- Stay under the long-context threshold when you can. Crossing it reprices the entire request, not just the overflow.
- Use Batch for overnight or bulk text jobs on models that offer the discount.
- Don’t send Priority or US-regional traffic unless you need the latency or the US processing guarantee; both cost more than the global standard endpoint.
- Call Grok 4.7 for text, and the dedicated Imagine or Voice models only when you actually need media.
- Set auto top-up caps and watch the usage explorer so a loop cannot drain an uncapped invoice limit.
For how other labs present token prices, see OpenAI API pricing and Gemini API pricing.
FAQ
Does SuperGrok or The X Subscription include Grok API usage?
Paid plans include Grok API access, but API usage is billed separately at standard model rates and does not draw from the plan's usage pool. That wording is on x.ai/pricing. You still need prepaid credits or invoiced billing in the xAI console.
Is the Grok API billed per token?
Text models are billed per million input, cached input and output tokens. Some models switch to a higher long-context rate once the prompt reaches the listed threshold, and that higher rate applies to all tokens in the request. Image and video models are billed per image or per second; Voice has its own per-minute, per-hour or per-character rates. Live numbers are on xAI's pricing and models pages.
How do prepaid credits and invoiced billing differ?
Prepaid credits are the usual path: you buy credits, and each request deducts from the balance. Monthly invoiced billing is off by default; if you raise the invoiced limit above zero, usage beyond prepaid credits is billed at month-end. If the invoiced limit stays at zero and credits run out, requests are rejected.
Can I use Grok 4.7 Fast on the public API?
No. xAI's pricing page says Grok 4.7 Fast is the same model on faster infrastructure, available only in Cursor and Grok Build, billed through your plan there. It is not available on the public xAI API, and Grok Build's free tier does not include it.
Are prepaid Grok API credits refundable?
xAI's billing FAQ says it cannot offer refunds on prepaid credit purchases unless required by law. It also cannot retroactively regenerate an invoice after you change billing details, so check the invoice information before you pay.
Sources
- xAI (SpaceXSI) Docs: API pricing (accessed 2026-11-18)
- xAI (SpaceXSI) Docs: Models (accessed 2026-11-18)
- xAI (SpaceXSI) Docs: Manage billing (accessed 2026-11-18)
- xAI (SpaceXSI) Docs: API billing FAQ (accessed 2026-11-18)
- x.ai: Pricing (The X Subscription and API FAQ) (accessed 2026-11-18)