Short answer
I can’t confirm a publicly documented production API for Grok 4 Heavy. xAI’s Grok 4 announcement describes Grok 4 as available through the API, but identifies Grok 4 Heavy with the SuperGrok Heavy tier. The API documentation reviewed here provides examples for other named models, including grok-4.7, but does not establish a Heavy-specific model ID or request. Heavy API availability is therefore unconfirmed—not proven unavailable.
Don’t guess a model name such as grok-4-heavy or substitute grok-4.7 and call it Heavy. To check what your API key can use, call the official GET /v1/models endpoint with the API key and the API base URL for the environment you intend to use. The response lists model IDs available to that authenticating key. The models page in the console also shows the current model list. A model listed publicly does not by itself confirm that your key has access; check the list for your key. If Heavy is absent or a request is denied, ask xAI API support specifically whether Grok 4 Heavy has a production API model ID and whether your team is entitled to use it. Include your team ID and the model name when asking support, as the documentation’s error guidance requests. (Models endpoint; regional endpoint guidance)
What is documented for API integration
The official API quickstart documents creating an account and API key, setting the API base URL to https://api.x.ai/v1, and sending a request to the Responses API with a named model. Its example uses grok-4.7—not Grok 4 Heavy. The REST reference documents bearer-token authentication for inference requests. (API quickstart; REST API reference)
If grok-4.7 meets your requirements, this is a documented example using the OpenAI Python SDK:
pip install openai
export XAI_API_KEY="your_api_key"
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["XAI_API_KEY"],
base_url="https://api.x.ai/v1",
)
response = client.responses.create(
model="grok-4.7",
input="Reply with the word ready.",
)
print(response.output_text)
A successful call should return generated text, which this example prints as response.output_text. This verifies a request to grok-4.7, not to Grok 4 Heavy. (API quickstart)
Choosing what to do next
- You specifically require Grok 4 Heavy: Don’t ship an integration based on a guessed model ID. Check
GET /v1/modelswith the API key and API base URL you plan to use. If Heavy is not listed or the request is denied, ask API support whether a Heavy API model ID exists and whether your team has access. A subscription tier or a public model listing alone does not confirm API-key entitlement. (Models endpoint; regional endpoint guidance) - You need a documented Grok API integration now: Test a model explicitly named in the official API materials, such as
grok-4.7, and evaluate it against your workload. Treat it as an alternative, not as an equivalent or alias for Heavy. (API quickstart) - You’re comparing costs: The official
grok-4.7model page lists US$2 per 1 million input tokens and US$6 per 1 million output tokens; cached input tokens are listed separately at US$0.50 per 1 million tokens. These aregrok-4.7rates, not Grok 4 Heavy pricing. (grok-4.7model page)
The November 2024 xAI API announcement is historical launch material: it discusses the grok-beta preview and a time-limited beta credit offer, not Grok 4 Heavy. It should not be used as current Heavy integration or pricing guidance. (2024 API announcement)
Sources
Models | Inference API — REST API Reference