Grok is a family of large language models made by xAI, the AI company founded by Elon Musk in 2023. You will most often meet it inside X (formerly Twitter), where it answers questions, summarises threads, and drafts posts. Developers can also call the same models directly through the xAI API.
Where Grok lives
There are two distinct products under one name. The consumer chat experience sits inside the X app and on the standalone Grok site and apps. The API is a separate pay-per-token service at api.x.ai, with its own console and billing.
That split matters for cost. A person chatting on X pays through a subscription (or uses the free allowance). A developer pays per token, and the price depends heavily on which Grok model they pick. We cover both in Lesson 3.
What Grok is good at
- Live context from X. Grok can pull in what people are posting right now, which makes it unusually useful for "what is happening with X today" questions.
- Very large context windows. Several Grok API models list context windows of 500,000 to 2,000,000 tokens in our catalogue, which is enough to load whole codebases or long document sets in one go.
- A cheap fast tier. Grok 4 Fast is priced at the low end of the market, so high-volume classification, extraction, and summarisation jobs can run for very little.
- OpenAI-compatible API. If you already have code written against OpenAI's client library, switching the base URL is usually enough to try Grok.
Honest limits
Grok is not magic and it is not always the cheapest or best option. The full-size models (Grok 4, Grok 4.5, Grok 4.6) cost several times more than Grok 4 Fast, and for many everyday tasks you will not notice the difference in quality. Real-time X data is a strength only when the question is about X; for general research it can also pull in rumours and jokes.
Like every model in this class, Grok can state wrong facts confidently, invent citations, and drift on long multi-step tasks. Treat its output as a draft to check, not an answer to trust. The consumer product also changes often, so features you read about last month may have moved or been renamed.
How it compares
Against ChatGPT and Claude, Grok's clearest differences are the X integration and the aggressive pricing on its fast tier. Its clearest weakness is a smaller ecosystem of tools, plugins, and third-party integrations. For a broader view see best ChatGPT alternatives and the cheapest AI tools roundup.
Knowledge check
Which statement about Grok is correct?