What Is Grok? xAI's AI Assistant Explained
Quick answer
Grok is a family of large language models built by xAI. You can use it at grok.com, in the iOS and Android apps, or inside X, and developers can reach the same models through the xAI API. It answers questions, writes and reviews code, and can search the web or X when you turn those tools on.
Grok is the model; Grok bots are what people build on top of it. That distinction is the source of most of the confusion, so this guide starts with the model itself — what it is, where you use it, and what it can and cannot do — before connecting it back to the bots in this directory.
The one-paragraph definition
Grok is a family of large language models developed by xAI. It is the assistant you reach at grok.com, in the official iOS and Android apps, and inside X — the same underlying models, different front doors.
For developers, the xAI API exposes those models programmatically at api.x.ai/v1. The documentation notes the API is compatible with the OpenAI and Anthropic SDKs, so switching from another provider is largely a change of base URL, key and model name.
Where you can actually use it
There are four routes, and they are separate products rather than one account with four screens. Grok.com is the standalone assistant. The iOS and Android apps mirror it with the conversations, settings and subscription syncing across platforms once you sign in.
Grok in X is the third route: fewer features than the standalone app, but tighter integration with X itself — reading and analysing posts is the point there. The fourth is the xAI API, which is what you use to build something rather than to chat.
One practical consequence worth knowing: xAI's own documentation states that a purchase on X, such as X Premium, does not affect your service status on the xAI API, and vice versa. They are billed and provisioned separately.
The models underneath
Grok is not a single model. xAI's documentation lists a flagship reasoning and coding model, alongside dedicated models and APIs for images, video and voice.
The flagship at the time of writing is listed with a 500,000-token context window and configurable reasoning — the reasoning setting is the one that most changes both the quality and the cost of a request.
For anything beyond text, xAI points to separate products: its Imagine API for image generation and editing, and a video API. Treating Grok as a text model with bolt-on media generation is roughly how the documentation describes it.
What it cannot do by itself
The most useful limitation to understand: Grok has no knowledge of current events or data beyond what was present in its training data. That is stated plainly in xAI's documentation.
To get real-time information you enable server-side search tools — Web Search and X Search. This matters for how you prompt: if you ask about something recent without those tools on, a confident-sounding answer can be built on nothing. Turning them on is the difference between a model recalling and a model looking something up.
The second limitation is structural rather than technical: a general assistant is not configured for your specific task. That gap is exactly what Grok bots exist to close.
How Grok bots fit in
A Grok bot is a saved set of instructions running on top of Grok. The creator writes plain-English instructions defining a persona and its rules, and the result has its own link anyone can open and use immediately.
So: Grok is the intelligence, and a bot is a reusable configuration of it. If you are asking what Grok is, you probably want the assistant. If you keep re-explaining the same task to it, you want a bot — and that is what you will find throughout this directory.