Grok API (October 2026): Pricing, How to Get an API Key, Models, Imagine and Code Examples

Prices and limits below are from xAI's documentation at docs.x.ai on October 17, 2026. xAI now operates as SpaceXSI; the API, console and docs domains haven't changed. We checked the request formats and error messages ourselves with test keys.

Grok API pricing

Text models

Prices per million tokens, in USD:

The 200K rule: when a prompt reaches 200,000 tokens, every token in that request (input, cached input and output) is billed at double the rates above. A 199K-token prompt with a 2K-token answer on Grok 4.7 costs about $0.41. At 200K tokens the same request costs about $0.82. If you work with long documents, split or trim them to stay under the line.

Reasoning models also produce reasoning tokens, which xAI lists as a separately billed token type alongside input, completion, image and cached tokens. Budget for them the way you budget for output.

Grok Imagine: images and video

xAI recommends grok-imagine-image-2.0 for images and grok-imagine-video-1.5 for video. A 10-second clip costs $0.20 on 1.5 Lite and $0.80 on 1.5.

Tools, batch and priority

• Server-side tools are billed per call on top of tokens: web search $5 per 1,000 calls, code execution $5 per 1,000, collections search (RAG) $2.50 per 1,000, and X search $5 per 1,000 posts or $10 per 1,000 profiles returned.
• Batch API: 20% off for grok-4.3 and the grok-4.20 models. Grok 4.7, 4.6, 4.5 and Build get no batch discount. Image and video jobs in a batch are billed at standard rates.
• Priority processing costs 2x standard rates.
• US regional endpoint (https://us.api.x.ai/v1) keeps processing in the United States. It serves only grok-4.7 and grok-4.6, no image, video or voice, and costs 10% more.

How much do 1,000 tokens cost?

On Grok 4.7, 1,000 input tokens cost $0.002 and 1,000 output tokens cost $0.006. A typical chat turn with a 2,000-token prompt and a 500-token answer costs about $0.007, so 1,000 such requests cost about $7. On Grok 4.3 the same 1,000 requests cost about $3.75.

Cost calculator

Estimate a monthly bill at xAI's October 2026 list prices. The calculator prices reasoning tokens like output, so add them to the output field for a conservative estimate.

Grok API vs ChatGPT, Claude and Gemini APIs

Standard input / output prices per million tokens for prompts under 200K tokens. Prices for the other providers are from our OpenAI, Claude and DeepSeek pricing pages and Google's pricing page, checked October 15–17, 2026.

Grok 4.7 matches GPT-6.1 Sol and Claude Sonnet 5.5 on input and costs 40% less on output. For high-volume simple tasks, GPT-6 Luna and Claude Haiku 5.5 are much cheaper than any Grok model. Token counts differ between tokenizers, so compare on your real prompts.

How to get a Grok API key

1. Sign up at console.x.ai.
2. Add credits. xAI's quickstart tells you to load the account with credits before you start, and requests fail without them.
3. Open API Keys in the console and create a key.
4. Save it as an environment variable:

export XAI_API_KEY="your_api_key"

On Windows PowerShell, use setx XAI_API_KEY "your_api_key" and open a new terminal.

Is there a free tier?

Not today. xAI's pricing and quickstart pages list no free API tier, and you prepay credits before calling the API. The X Subscription plans (Free, Plus $8, Premium $30, Super $100, Ultra $200 a month) include usage inside the Grok apps and Cursor, and their documentation doesn't describe developer API credits. Older Reddit threads about free Grok API credits describe programs that aren't in xAI's current documentation.

Quickstart code

curl

curl https://api.x.ai/v1/chat/completions \
  -H "Authorization: Bearer $XAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "grok-4.7", "messages": [{"role": "user", "content": "Explain what an API rate limit is in one sentence."}]}'

Python with the OpenAI SDK

import os
from openai import OpenAI

client = OpenAI(api_key=os.environ["XAI_API_KEY"], base_url="https://api.x.ai/v1")

resp = client.chat.completions.create(
    model="grok-4.7",
    messages=[{"role": "user", "content": "Explain what an API rate limit is in one sentence."}],
)
print(resp.choices[0].message.content)
print(resp.usage)  # cached tokens appear in usage.prompt_tokens_details.cached_tokens

xAI also supports the Responses API (client.responses.create(model="grok-4.7", input="...")), which is what its own quickstart uses. Other SDKs: npm install ai @ai-sdk/xai for the Vercel AI SDK, and pip install xai-sdk for xAI's own Python SDK.

Generate an image

curl https://api.x.ai/v1/images/generations \
  -H "Authorization: Bearer $XAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "grok-imagine-image-2.0", "prompt": "A red bicycle leaning on a brick wall, film photo", "n": 2, "aspect_ratio": "16:9"}'

The response contains image URLs. They're temporary, so download the files promptly or ask for base64 output. Optional parameters include aspect_ratio (default auto), resolution (1k by default, or 2k) and quality (low, medium or auto).

Get more cache hits

Prompt caching is automatic: when consecutive requests start with the same messages, the matching prefix is billed at the cached rate. xAI recommends sending an x-grok-conv-id header with a stable ID per conversation to raise the hit rate. Keep system prompts and tool definitions at the start and unchanged between turns.

Rate limits

Limits are per team and per model, in requests per second (RPS) and tokens per minute (TPM). Your tier depends on cumulative API spend since January 1, 2026, and never goes down:

Image models start at 6 requests per second and video models at 10 at Tier 0. Exceeding a limit returns 429 Too Many Requests. Your current tier and limits are on the Models page of the xAI console.

Errors you'll actually see

We sent requests with a wrong key and with no key to see the real responses:

Note the first row: xAI's debugging page lists 401 for invalid tokens, but a wrong key actually returns 400. With the OpenAI Python SDK that's raised as openai.BadRequestError, not AuthenticationError, so catch both if your code handles bad keys. For outages, check status.x.ai.

Using Grok through a multi-model gateway

If Grok is one of several models you call, a gateway lets you use one key and one bill for all of them. api.foo, the service that publishes these docs, serves Grok text models such as grok-4.7, grok-4.3 and grok-build-0.1, plus grok-imagine-image and grok-imagine-image-quality, at the same OpenAI-compatible paths (https://api.foo/v1/chat/completions, /v1/responses and /v1/images/generations). We confirmed each path reaches the gateway with a test key. Model names and prices are on the api.foo pricing page.

If you only use Grok, xAI's own API is the direct route and gives you the batch discount, regional endpoint and server-side tools.

Related

• Claude Code vs Codex vs Cursor vs OpenCode
• OpenAI API pricing and calculator
• DeepSeek API pricing and calculator
• What is an OpenAI-compatible API?
• API errors: 401, 429 and how to fix them

Sources

xAI (SpaceXSI) documentation, checked October 17, 2026: Pricing, Models, Quickstart, Rate limits, Prompt caching, Regional endpoints, Image generation, Debugging errors and The X Subscription pricing.