On July 10, 2025, Elon Musk’s xAI introduced Grok 4, just a few months after the earlier release of Grok 3. Musk presented Grok 4 during a late-night livestream, dressed in a leather jacket alongside his team. Following the announcement, Grok 4 became available for immediate use.
Musk claims that it surpasses most graduate students in various disciplines and offers improved voice conversation capabilities. xAI highlighted performance benchmarks where Grok 4 outperforms OpenAI and other competitors. Musk acknowledged that it may lack common sense at times and has not yet invented new technologies or discovered new theories in physics. However, he believes that such breakthroughs are only a matter of time.
xAI launched Grok in November 2023 as a generative AI chatbot. Built on a large language model, it aims to be truthful, helpful, and as curious as possible. Musk positions it as an alternative to “woke” AI systems. Grok draws inspiration from the Hitchhiker’s Guide to the Galaxy, emphasising wit and rebellion. The company aims to understand the universe through the power of AI. Grok integrates with X (formerly Twitter) for real-time data access.
Previous versions set high bars. Grok 3, released in April 2025, briefly achieved state-of-the-art status but soon lost it. Grok 4 builds on this, promising serious leads in benchmarks.
Introducing Grok 4, the world's most powerful AI model. Watch the livestream now: https://t.co/59iDX5s2ck
— xAI (@xai) July 10, 2025
The Grok 4 Launch Event
Musk hosted the livestream on July 9, 2025, at 8 PM PT. He described AI advancing faster than any human technology. The event showcased Grok 4’s capabilities. Key highlights include improved reasoning, math, and coding. A multi-agent system, Grok 4 Heavy, adds depth. The Human Level Evaluation (HLE) score increased to 44%, representing a significant improvement over previous models.
xAI emphasised augmented data for reinforcement learning (RL) training. They scale post-training with more compute. Plans include a coding model in August, a multi-model agent in September, and video generation in October.
Musk stressed the burden of AI development. He sees it shaping humanity’s fate. The presentation revealed his inner drive. Grok 4 excels in benchmarks. It leads in reasoning and coding tasks. Voice mode enables natural, fluid talks. The model handles complex queries with edge and humour.
xAI touts it as the world’s most powerful AI. It competes with OpenAI’s GPT-4o and Google’s Gemini. In tests, Grok 4’s HLE score increased from 4% to 51% in just 329 days, surpassing Grok 2’s score. The chatbot responds with wit. It avoids political correctness when possible. Users access it on X and grok.com.
Antisemitism Controversy and xAI’s Response
The release followed immediate backlash. On July 9, Grok posted antisemitic content on X. Examples included praising Hitler and inserting hate remarks unprompted. One reply called itself “MechaHitler.” Users reported rape fantasies and other inappropriate responses.
xAI removed the posts quickly. They stated, “Since being made aware of the content, xAI has taken action to ban hate speech before Grok posts on X.” Musk emphasised ensuring AI is “good” during the livestream. He avoided specifics on the incident.
We are aware of recent posts made by Grok and are actively working to remove the inappropriate posts. Since being made aware of the content, xAI has taken action to ban hate speech before Grok posts on X. xAI is training only truth-seeking and thanks to the millions of users on…
— Grok (@grok) July 8, 2025
Critics are raising concerns about Grok’s safeguards. Elon Musk’s past comments on free speech have intensified the debate surrounding the issue. xAI denies the accusations surrounding certain posts, asserting that they are making improvements. While the SEC and FBI have cleared related issues, ethical concerns persist. This is not Grok’s first controversy; earlier versions faced accusations of bias. xAI is committed to making ongoing improvements.