Elon Musk unveiled Grok 4, the latest version of his artificial intelligence chatbot, during a livestream on Wednesday night, just one day after the previous version, Grok 3, published antisemitic content on X, including a post that praised Adolf Hitler. Musk described the new model as "the smartest AI in the world," showcasing its performance on advanced math problems, scientific visualizations, and predictive tasks.
"Grok 4 is smarter than almost all graduate students in all disciplines simultaneously," Musk said during the event. "It is post-graduate, Ph.D.-level in everything." The demo featured Grok solving complex equations, generating an image of colliding black holes, and estimating the Dodgers have a 21.6% chance to win next year's World Series.
The release comes amid mounting controversy over Grok 3, which was pulled offline on Tuesday after users prompted it to produce Holocaust denial content and antisemitic rhetoric. Musk and xAI attributed the incident to the chatbot's excessive willingness to follow user instructions. "Grok was too compliant to user prompts. Too eager to please and be manipulated, essentially. That is being addressed," Musk wrote on X.
In a separate statement, xAI said, "We are aware of recent posts made by Grok and are actively working to remove the inappropriate posts. xAI has taken action to ban hate speech before Grok posts on X." The company added that it is "training only truth-seeking" and relies on feedback from "millions of users" to improve the model.
Despite the controversy, Musk pushed forward with the product rollout. Grok 4 introduces a new "heavy" mode, a high-performance tier using multiple AI agents to solve complex problems. That version is priced at $300 per month. The updated system also includes a voice interaction mode that Musk described as more responsive and natural, though it stumbled during a live demo when asked to compose an opera about Diet Coke.
Grok 4 has been benchmarked against leading industry metrics and reportedly topped "Humanity's Last Exam," a test composed of highly advanced problems designed to assess AI intelligence. Musk told viewers the rapid evolution of AI remains "terrifying" in its pace.
While Grok 4's technical enhancements drew praise from Musk and xAI researchers, the team avoided any mention of Tuesday's viral meltdown during the presentation. The omission drew criticism from some AI observers who noted the timing of the launch appeared to sidestep accountability.
xAI also announced upcoming releases for the fall, including a new coding model in August, a multimodal agent in September, and a video-generation system in October, signaling an aggressive development schedule despite the platform's ongoing moderation challenges.