What is Grok? Grok is the AI chatbot from xAI, Elon Musk’s AI company, which answers questions and generates text, code, images and video. Its biggest difference from other chatbots is the direct link with the X platform, which lets Grok retrieve real-time information from the web and from X.
Grok is a form of generative AI and basically works like other large language models. It recognises patterns in huge amounts of text and generates an answer based on them. What sets Grok apart is its real-time data and a deliberately more irreverent tone.
What is Grok?
What is Grok, in short? An AI assistant from xAI. You can use Grok through its own app, via grok.com and directly within X. Grok was launched in November 2023 and works much like ChatGPT, except that it has real-time access to online information. You use it for answering questions, writing, programming, searching and creating images and audio.
On 12 October 2026, SpaceXAI announced The X Subscription, a single subscription for Grok, Grok Bot, Cursor and X Premium. Alongside a free tier there are paid tiers at $8 (Plus), $30 (Premium), $100 (Super) and $200 (Ultra) a month, which xAI shows in dollars and excluding taxes. New customers can no longer sign up for SuperGrok and X Premium. For most tasks you can stick with the standard version. The heavier variants are worth it for complex reasoning and larger documents.
Who developed Grok?
Grok was developed by xAI, the AI company Elon Musk founded in March 2023. Musk was previously a co-founder of OpenAI, the company behind ChatGPT. After a disagreement about its direction, he left OpenAI in 2018 and later started xAI, with the aim of building an AI that, in his own words, puts truth and objectivity first.
In March 2025, xAI took over the X platform in an all-share deal. In February 2026, SpaceX then took over xAI itself, bringing Grok, X and Musk’s space business under one company; the AI division has since continued as SpaceXAI. Those takeovers explain the close integration between the chatbot and the social network X.
Where does the name Grok come from?
The name Grok comes from Robert A. Heinlein’s 1961 science fiction novel Stranger in a Strange Land. In that book, the verb “grok” means to take something in so completely that you see through it intuitively.
xAI chose the name deliberately, as a reference to that idea of deep comprehension. The name says nothing about the actual capabilities. A language model recognises statistical patterns and does not “understand” in a human sense.
How does Grok work?
At its core, Grok is a large language model trained on huge amounts of text that can also retrieve real-time information on top of that. Two things make it different from a standard chatbot: the link with live data and an adjustable reasoning mode.
With the DeepSearch feature, Grok searches the current web and the stream of posts on X. That lets it respond to trending topics and recent news, where models without live access are stuck with their training data.
Think Mode, also called Big Brain Mode, has Grok break a question down into steps before it answers. For complex questions or questions with several steps, that usually produces more accurate answers than a direct response.
What features does Grok have?
Grok’s features cover text, images, video and speech. Besides answering questions, you use it for writing, programming, searching and creating visual content.
Generating text and code
For text and code, Grok works as you would expect from a chatbot. You ask a question or give an instruction and get back an answer, a draft or a piece of code. Recent versions also produce complete documents, spreadsheets and presentations from a conversation.
Creating images and video
For images and video, Grok uses the Imagine feature, which lets you generate images and create short video clips from a text description. That distinguishes Grok from chatbots that only produce text.
Speech with Grok Voice
With Grok Voice, you have spoken conversations with the assistant instead of typing. The feature is available in the mobile app and has also been rolled out in Tesla vehicles. When I tested Grok, the voice mode proved handy for short questions on the go, less so for longer, substantive tasks.
Which Grok models are there?
Grok models appear at a rapid pace, with several versions a year; the overview of Grok models lists them all. The newest model is Grok 4.7, released on 21 September 2026. Before that, Grok 4.5 appeared on 16 July 2026, developed together with Cursor, the coding tool SpaceX has acquired. xAI positioned Grok 4.5 as comparable to Claude Opus, but faster and cheaper per token; the company’s own benchmarks support that claim only in part.
Grok 4.6 followed on 12 August 2026, with the emphasis on agents that keep working independently for long periods and on interactive and visual work. SpaceXAI calls Grok 4.7 its strongest model for coding and knowledge work, twice as fast at half the price of comparable models; that claim comes from the provider itself. Grok 4.5 was not immediately available in the European Union at launch; SpaceXAI announced the European roll-out for a later date.
According to xAI, the newest Grok model, Grok 4.7, is available in Cursor, in Grok Build and through the Grok API and cloud platforms; the announcement does not mention the Grok app. Grok Build is xAI’s coding agent and has been open to every subscription on web and mobile since 19 August 2026.
There is also a Heavy variant within the most expensive subscription, aimed at the most complex tasks. For everyday use the standard version is enough, while the heavier models only pay off for intensive reasoning work or large documents. xAI releases new versions at a high pace, sometimes two model generations within a few weeks.
How does Grok differ from other AI chatbots?
Grok differs from other AI chatbots mainly through its real-time link with X. Where Anthropic’s Claude and Google Gemini rely on their training data and separate search features, Grok pulls posts and discussions directly from the live stream on X.
That makes Grok attractive if you follow current events on social media, but it is not an automatic winner. Which model fits best depends on your task. The difference between Grok and ChatGPT lies both in the models themselves and in the type of question you ask.
When can you not trust Grok’s output?
You cannot take Grok’s output on trust, any more than the output of any other chatbot. A language model generates the most likely answer, not necessarily the correct one, and it can make up facts, figures or sources that sound plausible.
With Grok, two further points call for care. The real-time link with X means that unverified posts and rumours can end up in an answer. Your input may also be processed via the platform, which raises privacy questions. Its deliberately looser moderation led to investigations of Grok by regulators in 2026, including in the European Union.
Based on my evaluation of Grok compared with other LLMs, the advice is the same as for any AI tool: use the output as a draft, not as a finished product. In my work at the major banks, I saw how valuable a fixed checkpoint is, and that four-eyes principle translates directly to AI, with someone checking the output before it goes anywhere.
For teams, individual use is only half the picture: you also need a shared way of working that sets out such checkpoints. In the AI Literacy Course, your team learns to work with AI according to the same agreements, including agreements on when output is and is not usable.
Grok is a serious AI chatbot with its own niche in real-time information, although your way of working determines whether it adds value. At LearnLLM you learn to use AI tools responsibly and effectively, with attention to exactly those moments when you had better not trust the output.
Frequently asked questions about Grok
If you use Grok within X, you will come across data-processing settings that you manage for each account.
How do you stop X from using your posts to train Grok?
You stop X from using your posts to train Grok via Settings, Privacy & Safety, Data sharing and personalization and then Grok & Third-party Collaborators. Under Data Sharing, untick the option to share your public data and your interactions. According to the X Help Center, the Protect your posts setting also keeps your posts out of training and out of Grok’s answers. X can still use thumbs-up or thumbs-down feedback for training.
How do you delete your conversation history with Grok on X?
You delete your conversation history with Grok on X via Privacy & Safety, Data sharing and personalization and then Grok, using the Delete Conversation History option. According to the X Help Center, deleted conversations disappear from X’s systems within 30 days, unless they are kept longer for security or legal reasons. The instructions in the Help Center cover Grok within X, not the separate app or grok.com.
Is Grok available in the EU?
Yes. According to the X Help Center, Grok works in all countries where X Premium is available, which includes the Netherlands, so you can check availability in your country by seeing whether X Premium is offered there. You use it within X on iOS, Android and the web. New models or features may appear later in the European Union than elsewhere, as happened with Grok 4.5.
