xAI Released "The Smartest AI in the World" — Grok 4

xAI, Elon Musk’s artificial intelligence startup, has unveiled Grok 4, which it calls “the world’s most powerful AI model.” Alongside the launch, the company introduced a new SuperGrok Heavy subscription plan priced at $300 per month. The entrepreneur announced that the new AI can solve complex engineering problems whose solutions are not available on the internet or in books. “Grok 4 is at a level where it almost never gets math and physics questions on exams wrong—except when they are intentionally designed with a trick. It can identify errors or ambiguities in questions and either correct them or provide answers for all possible interpretations,” Musk said. During a live broadcast, Musk claimed that the new chatbot surpasses the level of a PhD in all subjects. “It may lack common sense at times, and it has yet to invent new technologies or discover new physics, but it's only a matter of time,” Musk added. Additionally, xAI introduced Grok 4 Heavy, a multi-modal version of Grok with enhanced performance. According to Musk, the neural network runs multiple agents to solve a problem simultaneously, then compares their answers to select the best result. Grok 4 performed well in several benchmarks, including Humanity's Last Exam—a test evaluating AI's ability to answer thousands of user-generated questions in math, humanities, and science. The chatbot scored 25.4% on this exam, outperforming Google’s Gemini 2.5 Pro (21.6%) and OpenAI’s O3 (21%). In the ARC-AGI-2 benchmark, Grok achieved a new advanced score of 16.2%. This test includes puzzle tasks where the AI must recognize visual patterns. Grok 4’s Performance in Benchmarks Benchmark Grok 4 Score Gemini 2.5 Pro OpenAI O3 Humanity's Last Exam 25.4% 21.6% 21% ARC-AGI-2 16.2% — — SuperGrok Heavy subscribers will gain access to the high-performance version of Grok and early trials of xAI’s products in development, including: A programming model A multimodal agent An AI video generator The new subscription plan will also offer: Advanced reasoning capabilities Programming tools Prioritized technical support Increased usage limits Features such as DeepSearch, Grok Studio, and Big Brain xAI is releasing Grok 4 via API, enabling developers to build applications on top of it. During the presentation, the team demonstrated Grok 4’s capabilities. The model can recognize video games and assess their addictiveness, as well as analyze data from X (formerly Twitter) and make predictions on Polymarket. The release comes amid controversy, as Grok recently drew criticism for making contentious statements. In July, after an update, the chatbot became more categorical and started issuing controversial and contradictory responses. xAI later stated it was working to remove inappropriate outputs. Previously, Grok made comments about the “white genocide” in South Africa without user prompting and questioned the number of Jews who died in the Holocaust. This behavior was attributed to “unauthorized modification of the prompt.”