Elon Musk’s AI company Xai released its latest flagship AI model, the Grok 4, late Wednesday, and announced SuperGrok Heavy, an AI subscription plan that costs $300 per month.
Grok is Xai’s answer to models such as Openai’s ChatGpt and Google’s Gemini, allowing you to analyze images and respond to questions. Over the past few months, Grok has been deeply integrated into Musk’s social network X, and recently acquired by Xai. But it also places Grok’s fraud fronts and centers for millions of users.
There are high expectations for the Grok 4. XAI’s latest AI models are stacked against Openai’s upcoming AI model GPT-5, which is scheduled to be released later this summer.
“When it comes to academic questions, Grok 4 is superior to the PhD level in all subjects, with no exceptions,” said Elon Musk in a live stream Wednesday night. “Sometimes, it may be lacking in common sense, and we have not yet invented new technologies or discovered new physics, but it’s only a matter of time.”

The launch of Grok 4 is in a tumultuous week for Elon Musk’s company. Earlier on Wednesday, Linda Jaccarino resigned from her role as CEO of X after about two years at the company. X has yet to announce her successor.
Yaccarino’s departure comes days after Grok’s official official, automated X account, responded to users with anti-Semitic comments criticizing Hollywood’s “Jewish executives” and praise Hitler. Xai had to easily restrict Grok’s accounts and remove offensive posts. In response to the incident, it appeared that Xai had removed the recently added section from Grok’s public system prompt, a list of instructions that the AI chatbot should follow.
Musk and Xai leaders largely shunned discussion of the incident, instead focusing on the performance and capabilities of the Grok 4.
On Wednesday, Xai launched two models: the Grok 4 and Grok 4 Heavy, the “multi-agent version” of the company that offers improved performance. Musk argued that heavy spawning of Grok 4 would cause multiple agents to work simultaneously to work multiple agents at the same time, and then they all insisted on comparing their works “like a research group” to find the best answer.
The company claims that the Grok 4 has shown frontier-level performance on several benchmarks, including the final exam for humanity. This is a challenging test that measures the AI’s ability to answer thousands of crowdsourced questions on subjects such as mathematics, humanities, and natural sciences. According to Xai, the Grok 4 won 25.4% in the final exam of humanity without a “tool”, surpassing Google’s Gemini 2.5 Pro, which won 21.6%, and Openai’s O3 (High), which won 21%.
Xai claims that the Grok 4 Heavy with the “Tools” can achieve a score of 44.4%, surpassing the Gemini 2.5 Pro using Tools, which scored 26.9%.
The non-profit ARC award says GROK is achieving new cutting edge scores in the ARC-AGI-2 test. This is another difficult benchmark consisting of puzzle-like problems where AI needs to identify visual patterns, earning 16.2%. This is almost twice the score for the next best commercial AI model, the Claude Opus 4.

Alongside the Grok 4 and Grok 4 Heavy, Xai has launched its most expensive AI subscription plans. This launched a $300/month subscription called SuperGrok Heavy. Planning subscribers will get an early preview of Grok 4 Heavy and early access to new features. The plan is similar to the ultra-premium tier offered by Openai, Google and humanity, but Xai currently offers the most expensive subscription of the leading AI providers.
SuperGrok heavy subscribers may have early access to some new products that Xai is due to be released in the coming months. The company said Wednesday that AI coding models will arrive in August, a multimodal agent in September, and a video generation model in October.
Xai is releasing GROK 4 through APIs to let developers build applications using models. The company notes that Xai’s enterprise sector is only two months old. However, we plan to work with HyperSchool to make Grok available through our cloud platform.
Despite Grok’s frontier-level performance on benchmarks, it may prove difficult for Xai to move past the recent disaster as Xai tries to sell Grok to companies as a real candidate for ChatGpt, Claude and Gemini. It is still unclear whether companies are ready to adopt GROK, flaws, etc.
Source link