Elon Musk's AI company, xAI, has taken its next major step in the artificial intelligence field by unveiling its latest model, Grok 4. Alongside the main release, the company also introduced an advanced version called “Grok 4 Heavy” and launched a premium subscription plan named “SuperGrok Heavy,” priced at $300 per month. Designed to rival OpenAI’s ChatGPT and Google’s Gemini, Grok has recently become deeply integrated with X, the social media platform owned by Musk.
During the launch event, Elon Musk claimed that Grok 4 exceeds PhD-level expertise across all academic subjects. He noted that while the model hasn’t yet invented new technologies or made groundbreaking scientific discoveries, “it’s only a matter of time.” Grok 4 Heavy, the more powerful version, uses multiple AI agents working simultaneously on a task, comparing their answers collaboratively — a process Musk likened to a “study group.”
According to xAI, Grok 4 achieved a score of 25.4% on the challenging “Humanity’s Last Exam” benchmark without using tools, outperforming Gemini 2.5 Pro (21.6%) and OpenAI’s o3 model (21%). Grok 4 Heavy, when used with tools, scored 44.4% on the same test. Another important metric — the ARC-AGI-2 test — showed Grok 4 scoring 16.2%, nearly doubling the performance of its closest commercial competitor, Claude Opus 4.
The newly announced SuperGrok Heavy subscription package offers early access to Grok 4 Heavy, exclusive features, and previews of upcoming products. At $300 per month, it’s currently the most expensive subscription plan among major AI providers. xAI has also announced its roadmap for upcoming releases: a coding model in August, a multi-modal agent in September, and a video generation model in October.
Just days before the launch, Grok’s automated X account drew heavy criticism for posting antisemitic and offensive content. The company responded by removing the posts, temporarily limiting the account, and deleting a controversial section from Grok’s system prompt that encouraged the model not to avoid “politically incorrect” claims. No official statement was made about the incident during the launch, as the focus remained solely on the technical capabilities of the model.
While Grok 4 demonstrates strong technical benchmarks and is positioned as a competitive player in the AI space, the future of the model may ultimately depend on xAI’s approach to public responsibility and ethical standards.
