Elon Musk’s artificial intelligence (AI) company, xAI, has unveiled its latest AI chatbot, “Grok 3.” A live demonstration was conducted on February 18th (Korean Time) via X (formerly Twitter), where Musk claimed that Grok 3 surpasses existing AI models like OpenAI’s ChatGPT and Google’s Gemini in performance.
Grok3 is not merely an upgrade to an AI model. It features enhanced logical reasoning and computational abilities, along with extended memory and adaptability. Amid ongoing discussions about the potential for AI chatbots to replace search engines, Grok 3 targets the search market by incorporating Deep Search functionality. Musk expressed confidence, stating that it is “the most powerful AI chatbot in human history.”
What makes Grok 3 different? It utilizes over 10 times the computing resources compared to its predecessor. To achieve this, xAI built its own supercomputer, “Colossus,” and deployed over 100,000 NVIDIA H100 GPUs. The computational speed has drastically improved over Grok 2, providing more precise responses, according to xAI.
The learning approach has also been differentiated by utilizing synthetic datasets and incorporating self-error correction and reinforcement learning techniques. This design allows the AI model to solve problems independently and derive optimal answers, enhancing logical reasoning, computational ability, and adaptability more than previous models.
One of the most notable features is its ability to retain memory and context. Previous AI chatbots had issues losing context or failing to remember previous messages as conversations lengthened. However, Grok 3 is designed to maintain consistency even in prolonged discussions.
In benchmark tests, Grok 3 outperformed major competitor models such as GPT-4o, Gemini-2 Pro, and Claude 3.5 Sonnet across various fields, including math, science, and coding. In tests conducted on the AI performance evaluation platform “Chatbot Arena,” Grok 3 scored the highest points in key categories. It displayed a robust performance, especially in the domains of logical reasoning and problem-solving ability.
Musk emphasized that “Grok 3 outperformed all competing models, including GPT-4o, in specific tests,” forecasting a shift in the AI chatbot market’s landscape.
Another core change in Grok 3 is the enhancement of its search capability. While existing AI chatbots simply summarized and provided information, Grok 3 has taken a step further by integrating a “Deep Search” feature. Deep Search isn’t just a simple search function. The AI analyzes hundreds of sources and generates its own report based on reliable information. Unlike existing search engines that provide links, Deep Search distills the essence of the information users seek and offers it.
Musk remarked, “It’s a powerful tool that can replace Google Search,” indicating that it will also compete in the search market.
Currently, Grok 3 is available to X (formerly Twitter) premium plus subscribers. Additionally, through a separate subscription service called “Super Grok,” users can gain early access to more advanced features and new updates. A voice interaction feature is also in development; however, it was not disclosed during the recent demonstration due to stability issues. Musk stated, “The voice mode is somewhat unstable” and announced that it will be introduced next week.
Musk’s ambition in the AI market extends beyond simple chatbot competition. Through Grok 3, he aims to establish an independent AI ecosystem while reaching into the search market. Musk is attempting to evolve the X platform into a “Super App” by integrating AI. If a comprehensive service encompassing AI-based chatbots, search functionality, and voice assistants becomes possible, it could significantly alter the AI market, currently dominated by Google and OpenAI.
However, competitors aren’t remaining passive. Google, OpenAI, and Microsoft continue to improve the performance of their AI chatbots and are preparing next-generation models. Whether Grok 3 can maintain a sustained lead in the AI chatbot market or will only capture initial attention will depend on future performance validation and market responses.