xAI Unveils Grok 4: The World’s Most Intelligent AI

Elon Musk’s startup xAI has introduced Grok 4, touted as “the world’s most powerful AI model.” A new subscription plan, SuperGrok Heavy, is also launched at $300 per month.

Introducing Grok 4, the world’s most powerful AI model. Watch the livestream now: https://t.co/59iDX5s2ck

— xAI (@xai) July 10, 2025

The entrepreneur announced that the new artificial intelligence has managed to solve complex engineering problems, answers to which cannot be found online or in books.

Grok 4 is at the point where it essentially never gets math/physics exam questions wrong, unless they are skillfully adversarial.

It can identify errors or ambiguities in questions, then fix the error in the question or answer each variant of an ambiguous question. https://t.co/vB6NUOZTOX

— Elon Musk (@elonmusk) July 10, 2025

“Grok 4 is at such a level that it practically never makes mistakes on math and physics exam questions, except when they are deliberately tricky. It can identify errors or ambiguities in formulations and then either correct them or provide answers to all possible interpretations,” noted the billionaire.

During the livestream, he stated that the new chatbot surpasses the level of a PhD across all subjects.

“At times, it may lack common sense, and it has yet to invent new technologies or discover new physics, but it’s only a matter of time,” Musk added.

Additionally, the Grok 4 Heavy model was introduced—a multimodal version of Grok with enhanced performance. According to the head of xAI, the neural network launches several agents to solve a task simultaneously, after which they compare their answers to find the best result.

Grok 4 has shown high results in a number of benchmarks, including Humanity’s Last Exam—a test assessing AI’s ability to answer thousands of user questions in mathematics, humanities, and natural sciences. The chatbot scored 25.4% on this exam, surpassing Google’s Gemini 2.5 Pro (21.6%) and OpenAI’s o3 (21%).

In ARC-AGI-2, Grok achieved a new leading result of 16.2%. The test includes puzzle tasks where AI needs to recognize visual patterns.

In addition to access to the high-performance version of Grok, SuperGrok Heavy subscribers will be able to try the company’s products in development:

• a programming model;
• a multimodal agent;
• an AI video generator.

The new plan will also offer:

• enhanced reasoning capabilities,
• programming tools,
• priority technical support,
• increased usage limits,
• features like DeepSearch, Grok Studio, Big Brain.

xAI releases Grok 4 through an API. Developers can create applications based on it.

During the presentation, Grok 4’s capabilities were demonstrated. The model can recognize video games and assess their entertainment value, as well as analyze data from X and make predictions on Polymarket.

The release occurred amid a scandal involving Grok due to controversial statements.

Back in July, after a recent update, the chatbot became more assertive and began issuing contentious and contradictory statements. Subsequently, the xAI team stated that they are working on removing inappropriate publications.

Previously, Grok touched upon the topic of “white genocide” in South Africa without a corresponding user request and also questioned the number of Jewish deaths during the Holocaust. At that time, its behavior was explained as “unauthorized prompt modification.”