The world’s wealthiest man Elon Musk and a few members from his xAI team unveiled the latest iteration of its AI model – the Grok 4. The new version is a heavy update to its previous Grok 3 released earlier this year. Calling it the ‘smartest AI in the world’, the new AI model brings real-time data search, 10x more compute, new voices, enterprise grade security, frontier multimodal reasoning, and a lot more. Grok 4 has been made available for Premium+ users and aims to compete with leading AI models like OpenAI's ChatGPT and Google Gemini’s latest versions.
During the event, Musk refers to Grok 4 as ‘remarkably smarter’ than anything else. He says, “Grok 4 is smarter than almost all graduate students in all disciplines simultaneously.” Furthermore, Musk asserts that Grok 4 can reason at ‘superhuman levels’. Due to the massive improvements coming with the latest version, the xAI team said that it decided to skip the Grok 3.5 numerical and go directly for the Grok 4 version. Two versions of the Grok 4 were unveiled – Grok 4 single agent version and Grok 4 Heavy – a multi-agent enterprise version.
To know all of the new capabilities and details of the Grok 4, you can watch the livestream below:
What's new with Grok 4?
The Grok 4 benchmarks outperform all other AI models, as shown in the livestream, including the latest version of ChatGPT and Google Gemini. Grok 4 brings huge upgrades to voice conversations and introduces new voices, like Eve, as well. The new AI model offers a natural, human-like voice with fewer interruptions. The conversations are snappier, wittier, and the voices can sing as well. Grok 4 has been trained on xAI's Colossus supercomputer for advanced, scientist-grade reasoning.
The xAI team is offering Grok 4 to premium users only. The SuperGrok Heavy offers access to Grok 4 and Grok 4 Heavy for $300 per month, while SuperGrok offers access to Grok 4 and Grok 3 models both for $30 per month. The free plan currently offers Grok 3 access only. Few features like the coding model, the multi-model agent, and video generational model are coming in the next months.
