In streaming on X, Elon Musk’s xAI announced its new Grok-3 AI model, which has “more than 10 times” the computer power of its predecessor. According to a test conducted by the company, the new Grok-3 outperforms all the players, including DeepSeek’s V3, OpenAI’s GPT-4o, and Google’s Gemini.
In addition, xAI added to Grok-3 the magical word of the moment: Reasoning. Unlike the previous versions, the “reasoning models” are trained to answer complex questions or transform instructions into smaller tasks. These reasoning models also take a little longer to give users an answer so they can double-check the information and avoid AI hallucinations.
In total, xAI announced two Grok-3 reasoning modes and one Grok AI agent. According to the company, the new modes are “Think,” which will show Grok reasoning in real-time, and “Big Brain,” which will be especially useful for tasks that require more computational power. The AI agent, Deep Search, will be a “next-generation search engine,” analyzing information from the Internet and X to answer user queries.
Grok-3 is available for X Premium Plus subscribers. This new model launch also coincides with a new price hike for this subscription tier: $40, up from $22. Designed to be a “maximally truth-seeking AI,” xAI is also launching a new SuperGrok tier, costing $30 per month, offering “the most advanced capabilities and earliest access to new features.”
All the new features are available in the Grok app. xAI wants to add synthesized voice capabilities in the future and, like its competitors, open source previous versions of its AI models, such as Grok-2, so more people can take advantage of AI.
Since DeepSeek exploded in popularity last month, American companies have tried to catch up with the Chinese reasoning model and release more features faster. They’ve seen Eastern firms surpass the technologies they considered state-of-the-art.
BGR will keep reporting on this AI race as the companies announce new features and technologies.
