The xAI API: Building with Grok

Why Developers Choose xAI

The xAI API gives you programmatic access to the Grok model family through a REST interface that is compatible with the OpenAI Python and JavaScript SDKs. If you have already built with OpenAI's API, switching to xAI requires changing one line: the base URL. The same authentication pattern, the same message format, and the same function calling interface all work with xAI without changes.

What you gain by switching — or by adding xAI as a second provider — is real-time data. Every xAI API call has live X and web search built in. This means you can build AI applications that answer questions about what is happening right now without building a separate search integration. The context window is also large: Grok 4.3, the primary model, supports one million tokens.

Getting Your API Key

Visit console.x.ai — xAI's developer console. Create an account or sign in with your X credentials. Navigate to API Keys and create a new key. Copy it immediately; xAI does not show the full key again after creation.

All API requests authenticate with a Bearer token in the Authorization header:

Authorization: Bearer xai-your-api-key-here Base URL: https://api.x.ai/v1

The Model Lineup

As of mid-2026, the primary models available through the xAI API are:

Older model slugs (grok-4.1, grok-4-fast, grok-code-fast-1) were retired in May 2026 and now route to grok-4.3. Always check docs.x.ai for current slugs before publishing a production integration.

Your First API Call

A minimal chat completion call in Python using the OpenAI SDK:

from openai import OpenAI client = OpenAI( api_key="xai-your-key-here", base_url="https://api.x.ai/v1" ) response = client.chat.completions.create( model="grok-4.3", messages=[{"role": "user", "content": "What is happening in AI this week?"}] ) print(response.choices[0].message.content)

This is identical to an OpenAI call except for the base_url and api_key. Because Grok has live web search built in, the question "this week" will draw from current information — no additional configuration or search integration needed.

Advanced Features

• Function calling — Define tools and have Grok call them when it needs data. Works identically to OpenAI function calling, using the same tools array format.
• Streaming — Add stream=True for token-by-token output as it is generated. Identical to OpenAI streaming.
• System messages — Pass a system message in the messages array to set the model's persona, instructions, and constraints.
• Batch mode — Submit large volumes of requests at a reduced per-token cost. Useful for offline processing: document classification, bulk summarisation, large-scale data extraction.
• Server-side tools — Enable built-in tools like web search and code execution within the API call, without a client-side function calling loop.
• Long context — Send up to one million tokens in a single Grok 4.3 request. Useful for full codebase analysis, large document sets, or long conversation histories.

The Grok Imagine API

The xAI API also exposes image and video generation through the grok-imagine model endpoint. This lets you add Aurora's image generation and Grok Imagine's video capabilities to any application programmatically. Pricing is per image or video generated rather than per token. Check docs.x.ai for the current Imagine endpoint documentation and rate limits.

Try This

Set up a free account at console.x.ai, generate an API key, and run the minimal Python example above. Then modify the message to a time-sensitive question — something happening this week — and observe how live web access appears in the response. Next, add a system message: "You are a research assistant. Always cite your sources and note how recent each one is." Notice how the output structure changes. This three-step sequence confirms that the live search is working and that the API is responding to system-level instructions.