Grok

Grok

Developer:

xAI

Type:

generative chatbot

Initial release:

November 3, 2023

Interface language:

multilingual, including English

License:

proprietary; Apache-2.0 for Grok-1

Website:

grok.com

Grok is a generative artificial intelligence chatbot developed by xAI and launched in November 2023. It is integrated with the X social network, formerly Twitter, which allows it to draw on real-time information from that platform. At first, access to Grok was available only to subscribers of the highest X tier, Premium+.

Grok is known for an informal style and a “rebellious” tone in its answers. Its developers gave the chatbot a sense of humor, so it can answer with sarcasm and jokes. According to Elon Musk, Grok can answer almost any question, including “spicy” prompts that other chatbots reject. The name “Grok” is borrowed from Robert A. Heinlein’s science-fiction novel Stranger in a Strange Land, where it means deep intuitive understanding.

History

Grok chat window

Elon Musk was one of the co-founders of OpenAI in 2015, but left it in 2018 after disagreements over the organization’s commercial direction. Musk later criticized OpenAI models for what he described as excessive political correctness and said he intended to create his own “maximally truth-seeking” artificial intelligence. In July 2023, he founded a new company, xAI, announcing that its goal was to develop an independent chatbot based on large language models.

In early November 2023, xAI presented a prototype of the Grok chatbot to a limited group of users. Access to the beta version was initially offered only to some X users by personal invitation from Musk, and later to X Premium+ subscribers. The developers warned that Grok was at an early stage, describing it as a “very early beta product” trained for two months, and promised rapid improvement based on user feedback.

On March 11, 2024, Musk announced that the source code of Grok-1 would be opened, and six days later xAI released the Grok-1 model under the Apache 2.0 license, publishing its architecture and weights. In May 2024, an updated Grok-1.5 version was released with improved reasoning abilities and a larger context window, becoming available to all paid X subscribers.

On August 14, 2024, xAI released Grok-2 and the lighter Grok-2 mini, which received improved algorithms and a new text-to-image generation feature. The new models allowed users to create illustrations and publish them directly in the X feed, with very few content restrictions. In autumn 2024, Grok gained visual understanding for images and documents, while its web interface began supporting internet search and PDF uploads for analysis.

In February 2025, xAI released Grok-3, trained on new large-scale computing infrastructure. Musk called Grok-3 “the smartest AI on Earth”, pointing to significant progress in mathematical and scientific tasks compared with previous versions. Grok-3 showed higher results on a number of standardized tests and added support for voice mode, while its context window was significantly expanded for complex requests.

In July 2025, xAI introduced the flagship Grok-4 model together with a premium “SuperGrok Heavy” plan priced at US$300 per month. Grok-4 placed xAI at the top of several key AI benchmarks, ahead of models from OpenAI, Google and Anthropic. At the same time, Grok-4 received a record context window of 256,000 tokens and an improved multi-agent mode, which helped it surpass competitors in complex tests of knowledge and fluid intelligence.

Functionality

Grok functions as a virtual assistant with a broad range of capabilities. It supports conversational interaction, answering user questions and performing tasks in natural language, from writing program code to drafting text. The chatbot’s tone is intentionally informal: Grok can joke, use sarcasm and give “politically incorrect” comments, which distinguishes it from more restrained counterparts. According to Musk, the chatbot aims to provide maximally truthful and direct answers, even when they may seem provocative or uncomfortable.

Technically, Grok is based on xAI’s own large language models trained on extensive datasets. A distinctive feature of the service is access to up-to-date information: Grok can retrieve data from posts and publications on X in real time, supplementing its answers with fresh information from the internet. For example, the bot can scan recent news and social media on request to provide a current overview of events.

Modern versions of Grok have multimodal capabilities. The chatbot can process not only text prompts but also uploaded images, and can generate new images from text descriptions. Grok’s image-generation tool uses the Flux model from Black Forest Labs and allows generated images to be posted directly to X. Grok can also search the internet, summarize information it finds and work with uploaded documents, such as analyzing PDF files, broadening the set of tasks it can perform.

Availability

At launch, Grok was available only to subscribers of X’s paid Premium and then Premium+ plans. In December 2024, xAI opened free access to the chatbot for all X users: even without a subscription, users could send up to 10 prompts every two hours and generate up to three images per day. Paid subscribers retained advantages such as higher limits and early access to new bot features.

Over time, the ways to access Grok expanded. In January 2025, xAI released a standalone Grok mobile app for iOS, making it possible to use the chatbot outside X. A web version of the service, grok.com, also became available for use from any browser. In March 2025, an integrated Grok bot was launched in Telegram, further simplifying access to the AI assistant for a wider audience. Thus, by mid-2025, Grok could be used both inside X and outside the social network through mobile apps or third-party platforms.

Criticism

Grok has repeatedly been criticized for the lack of strict content moderation and for potential bias in its responses. In particular, after a system update in summer 2025, Grok posted a series of replies on X containing antisemitic statements and praise for Adolf Hitler, which caused a broad public reaction. xAI apologized for the messages, explaining the incident as a technical error in an update and removing system instructions that allowed “politically incorrect” responses. Observers noted that Grok’s focus on “maximum truthfulness” and free expression can lead to conspiracy theories, hate speech and other controversial content that is usually filtered in similar services. Some experts warn that the absence of strong moderation in such models attracts malicious users and complicates efforts to fight misinformation.

Prospects

Musk plans to deploy Grok technologies beyond internet services by integrating the AI assistant into Tesla products. In summer 2025, he announced that the Grok chatbot would appear in Tesla electric vehicles, where it could help drivers with voice requests and provide up-to-date information directly on the car’s display. According to Musk, combining Grok’s capabilities with hardware platforms such as cars and robotics, including the Tesla Optimus humanoid robot, could significantly expand the use of artificial intelligence in everyday life.

In late 2025, xAI launched Grokipedia, an experimental online encyclopedia generated by AI based on Grok models. Musk said that Grokipedia would be “orders of magnitude” better than Wikipedia in breadth and accuracy, but early reviews were skeptical. Analysis suggested that a large share of Grokipedia articles had effectively been copied from Wikipedia, while AI-added passages often lacked sources or contained errors. This highlighted the difficulty of creating a reliable knowledge base solely with AI and confirmed the dependence of such systems on verified content created by people.

In the future, xAI plans to continue developing the Grok model family and related services. The company promises specialized AI versions for narrower tasks, such as a coding assistant and video generation, along with regular improvements in response quality. Musk has also indicated an intention to open-source previous models after new generations are released, in order to encourage external work on fine-tuning and applications of Grok. According to observers, Grok’s further evolution will show whether xAI’s approach, with an emphasis on “truthfulness” and less censorship, can gain user trust and a place in the crowded AI chatbot market.