Intermediate

Grok API

The xAI API gives developers access to Grok models via a REST API at api.x.ai/v1. It is OpenAI-compatible — meaning existing OpenAI SDK code works with a base URL change.

Pricing (May 2026)

Model	Input / 1M	Output / 1M	Context	Notes
grok-4.3	$1.25	$2.50	1M tokens	Current flagship — best quality
grok-3-mini	$0.30	$0.50	131K tokens	Cheaper; chain-of-thought by default
grok-3 (legacy)	$0.30	$0.50	131K tokens	Legacy — prefer grok-3-mini
grok-2-vision-1212	$2.00	$10.00	32K tokens	Vision (image input) — legacy pricing

Free Developer Credits

$150/month in free API credits

xAI offers developers $150/month in free API credits through a data-sharing program. At Grok 4.3 prices ($1.25/$2.50 per 1M tokens), this covers around 50M input tokens or substantial experimentation. Sign up at console.x.ai to claim credits. This makes Grok one of the most cost-effective models to experiment with at real scale.

Quick Start

pip install openai

from openai import OpenAI

client = OpenAI(
    api_key="your_xai_api_key",
    base_url="https://api.x.ai/v1"
)

response = client.chat.completions.create(
    model="grok-4.3",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "What's trending on X right now?"}
    ]
)
print(response.choices[0].message.content)

API Features

OpenAI-compatible

Same request/response format as OpenAI. Change base_url and api_key — existing code works. Compatible with LangChain, LlamaIndex, and any OpenAI SDK.

Streaming

All models support streaming. Set stream=True for token-by-token output — reduces perceived latency in user-facing applications.

Tool / Function Calling

Grok 4.3 and Grok 3 Mini support tool calling — define functions in JSON schema, model decides when to invoke them. Use for building Grok-powered agents.

Live Search (X data)

Optional live_search parameter connects the model to real-time X/Twitter data during inference. Responses include cited sources from X posts.

Vision (image input)

grok-2-vision-1212 accepts image + text input. Send base64-encoded images or image URLs. Useful for image analysis tasks.

JSON mode

Force structured JSON output — useful for data extraction. Set response_format: {type: 'json_object'} in the API call.

Live Search — Real-Time X Data

response = client.chat.completions.create(
    model="grok-4.3",
    messages=[{"role": "user", "content": "What are people saying about the latest AI releases?"}],
    extra_body={"search_parameters": {"mode": "on"}}   # enable live X search
)
# Response includes citations from real X posts

When to Use the Grok API

Current events and trending topics — the only commercial API with native X/Twitter data access
Long document processing — Grok 4.3's 1M context window handles very large inputs
Cost-sensitive experimentation — $150/month free credits make it cheap to prototype
Alternative reasoning model — Grok 3 Mini uses chain-of-thought by default at competitive prices

Checklist: Do You Understand This?

Can you make a Grok API call using the OpenAI Python SDK?
Do you know the current flagship model's price and context window?
Do you understand how to enable live X search in an API call?
Can you claim the $150/month free developer credits?