Elon Musk's corporations xAI has launched its latest flagship AI model, Grok4. Similar to AI chatbots like OpenAI's ChatGPT, it can analyze images and answer questions.
According to IT news outlet TechCrunch, xAI revealed itself along with Musk through the social media platform X Live on the 9th (local time).
xAI claimed that Grok4 scored an accuracy rate of 25.4% on 'Humanity’s last exam,' regarded as a benchmark for AI performance, which is higher than Google's Gemini 2.5 Pro (21%) and OpenAI's o3 (21%). This exam is a kind of workbook created by the nonprofit AI Safety Center and AI startup ScaleAI, containing thousands of questions covering mathematics, physics, medicine, and the humanities.
On the same day, Musk said, 'Grok4 is at a Ph.D. level or higher in all subjects regarding academic questions,' noting, 'It may sometimes lack common sense and hasn’t yet discovered new technologies or physics theories, but that is just a matter of time.'
xAI also unveiled Grok4 Heavy, which has multi-agent capabilities. xAI explained, 'It generates multiple agents simultaneously to solve problems individually, then compares the results like a study group to derive the optimal answer.'
In addition, xAI announced it will launch a high-priced AI subscription service called SuperGrok Heavy, which costs $300 per month (approximately 400,000 won). High-priced plans from OpenAI, Google, and Anthropic are around $200 per month. SuperGrok Heavy subscribers will be able to experience Grok4 Heavy early and will also have access to future products from xAI.
Grok4 comes amid controversy after Grok repeatedly praised Adolf Hitler, the German dictator who orchestrated the genocide of Jews, and posted anti-Jewish content, leading to the announcement of Linda Yaccarino's resignation as CEO of X. Musk did not comment on this controversy that day.
xAI plans to release an AI coding model in August, a multimodal agent in September, and a video generation model in October.