The AI company in Elon Musk, Xai, on Wednesday releases the latest model of Flagship AI, Grok 4, and opened a new $ 300-monthly AI subscription.
Nrok Xai’s response to models like Openia’s ChatGPT and to Google’s Seminiand can analyze images and respond to questions. In recent months, the groke becomes more deeply united with the social network of Musk, X, which just got to xai. However, it is also placed The wrong grog and center attitude for millions of users.
Expectations are high for grok 4 GPP-5, which is expected to launch late summer.
“With regard to academic questions, the groke 4 is better than PhD level on each subject, no exception,” Elon Mok said on a lifstream Wednesday night. “Sometimes it is not a lack of common sense, and it has never invented new technologies or discovered new physics, but that’s just one hour.”
Groke 4 launch comes between a disturbed week for Elon Musk companies. Earlier on Wednesday, Linda Yackcarino went up from his role as CEO of X after two years in the company. X has not yet announced his successor.
Yacarino’s departure comes only days after the groc officer, automatically x accounts with users Antisemitic comments criticize “Hollywood Executives” and praising Hitler. Xai should turn out to short groke account and eliminate offensive posts. In response to the incident, Xai appeared to have Took a newly added section from Prompt Grek’s public systemA list of instructions for AI Chatbot to follow, saying this does not turn away from making “incorrect politics”.
Musk and Xai leaders are mainly avoiding reference to the event, however focusing on the performance and capability of grok 4.
Xai launches two models on Wednesday: Grok 4 and Grok 4 heavy – the last “multi-agent version” offering more performance. Musk grak claims 4 heavy spawns that many agents work in a problem simultaneously, and then they all compare their job “like a full answer.
Xai admits that the groke 4 appears ahead of frontier level, including the final example of the person who measures thousands of questions such as subjects such as mathematics, people, and natural science. According to Xai, the Grok 4 scored 25.4% of the human final exam “tools,” Pro, scoring 21.6%, scoring 21%.
Xai admits that the groke 4 is heavy, with “tools,” a score of 44.4%, outperforming Gemini 2.5 Pro with 26.9%.
Nonprofit arc prize says that Gro has achieved a new state-of-the-art score In Arc-agi-2 test – another difficult benchmark consisting of puzzle-like problems where AI should identify visual patterns – scoring 16.2%. That’s almost two times points to the next best model of commercial AI, Claude Opus 4.
With Grok 4 and Grok 4 heavy, Xai launches the most expensive subscription plan on AI, a $ 300-per-month subscription called Supergroc heavy. Plan subscribers get an early groke 4 preview, as well as early access to new features. The plan is similar to the ultra-premium tier offered by Openi, Google, and Anthropic, but Xai now offers the most expensive subscription to the major AI providers.
Supergroging Heavy Subscribers can get early access to some new Products XAI plans to launch the coming months. The company said Wednesday that a model of AI coding arrives in August, a multi-modal agent in September, and video model in October.
Xai releases the groke 4 by the API in an effort to get developers to build model applications. The company says the business sector in Xai only two months old, however, it plans to work in hypercals to use their platforms in the clouds.
Despite dealing with the Greek level level in benchmarks, xai can be proud to have the taking off the loss of chatgpt businesses, chatude, and Gemini. If the business adoption businesses are ready, errors and all, remain visible.
