According to xAI, Grok is significantly ahead of current competitors such as GPT-03 and Gemini 2.5 pro in key metrics. Musk claims that Grok 4 solves problems in any subject better than scientists with a PhD.
xAI has unveiled its new artificial intelligence models, the Grok 4 and Grok 4 Heavy. The latter is the company’s “multi-agent version” that offers increased performance. Musk says the Grok 4 Heavy creates multiple agents that work on a problem simultaneously, then they all compare their work “like a research team” to find the best answer.
Grok 4 was trained on xAI's Colossus supercomputer for advanced reasoning to achieve results at the level of leading scientists.
“When it comes to academics, Grok 4 is better than a PhD in every subject, bar none,” Elon Musk said during the presentation.
xAI claims that Grok 4 has demonstrated unprecedented performance in several tests, including Humanity's Last Exam, a complex test that measures AI's ability to answer thousands of questions in math, humanities, and science. According to xAI, Grok 4 scored 25.4%, ahead of Google's Gemini 2.5 Pro, which scored 21.6%, and OpenAI's o3, which scored 21%.
Introducing Grok 4, the world's most powerful AI model. Watch the livestream now: https://t.co/59iDX5s2ck
— xAI (@xai) July 10, 2025
The developers claim that Grok 4 will have a specialized model for Code programmers , reminiscent of competitor tools like GitHub Copilot or GPT-4 Code Interpreter.
Grok 4 is expected to support not only better text generation, but also images and possibly video, which Musk mentions as one of the model's biggest weaknesses. Grok 4 Voice has a natural, human-like voice with fewer pauses. Like previous Grok models, Grok 4 features DeepSearch, a tool that pulls real-time data from web resources, especially Musk's X platform, which will allow Grok to provide relevant results during chats.
Separately, the developers focused on understanding Internet culture. Grok 4 is tuned to accurately interpret memes, slang, and humor. The current version has already coped well with the meme potential of the Ukrainian X segment, entering into a skirmish with Russian users and bots.
xAI has launched its most expensive AI subscription plan, SuperGrok Heavy, priced at $300 per month. Subscribers to the plan will receive an early preview of Grok 4 Heavy, as well as early access to new features.
The announcement of Grok 4 took place against the backdrop of a scandal with the judgments of the current model, who allowed herself to write anti-Semitic posts, glorified Hitler, and openly attacked the leaders of many countries, including X CEO Linda Iaccarino, who resigned on the eve of the announcement of Grok 4.
