Grok, xAI's response to models like OpenAI's GPT-4o and Google's Gemini, is capable of analyzing images and answering questions, as well as powering several features in Musk's social network, X. Grok 3, which has been in development for several months, was optimistically slated for release in 2024 but missed that deadline.
To train Grok 3, xAI utilizes a massive data center in Memphis, housing around 200,000 graphics processors. In a post on X, Musk claimed that Grok 3 was developed with "10 times" the computational power of its predecessor, Grok 2, using an expanded dataset that allegedly includes materials from court cases.
"Grok 3 is orders of magnitude more powerful than Grok 2," Musk stated during a live presentation. "It's the most truth-seeking artificial intelligence, even if that truth sometimes contradicts what is politically correct."
More specifically, Grok 3 is a family of models. The smaller version, Grok 3 mini, answers questions more quickly but at the cost of some accuracy. Not all models and corresponding features of Grok 3 are available yet (some are in beta), but they began rolling out on February 17.
xAI claims that Grok 3 outperforms GPT-4o in benchmarks such as AIME (which assesses model performance on a sample of math questions) and GPQA (which evaluates models using PhD-level physics, biology, and chemistry tasks). According to xAI, an early version of Grok 3 also scored highly in Chatbot Arena—a crowdsourced test where various AI models compete against each other, and users vote for the best responses.