Musk has additionally apparently used the Grok chatbots as an automatic extension of his trolling habits, showing examples of Grok 3 producing “based mostly” opinions that criticized the media in February. In Might, Grok on X started repeatedly generating outputs about white genocide in South Africa, and most just lately, we have seen the Grok Nazi output debacle. It is admittedly tough to take Grok critically as a technical product when it is linked to so many examples of unserious and capricious functions of the know-how.
Nonetheless, the technical achievements xAI claims for numerous Grok 4 fashions appear to face out. The Arc Prize group reported that Grok 4 Pondering (with simulated reasoning enabled) achieved a rating of 15.9 p.c on its ARC-AGI-2 check, which the group says practically doubles the earlier business finest and tops the present Kaggle competitors chief.
“With respect to educational questions, Grok 4 is healthier than PhD degree in each topic, no exceptions,” Musk claimed through the livestream. We have previously covered nebulous claims about “PhD-level” AI, discovering them to be usually specious advertising speak.
Premium pricing amid controversy
Throughout Wednesday’s livestream, xAI additionally introduced plans for an AI coding mannequin in August, a multi-modal agent in September, and a video era mannequin in October. The corporate additionally plans to make Grok 4 obtainable in Tesla autos subsequent week, additional increasing Musk’s AI assistant throughout his numerous corporations.
Regardless of the current turmoil, xAI has moved ahead with an aggressive pricing technique for “premium” variations of Grok. Alongside Grok 4 and Grok 4 Heavy, xAI launched “SuperGrok Heavy,” a $300-per-month subscription that makes it the costliest AI service amongst main suppliers. Subscribers will get early entry to Grok 4 Heavy and upcoming options.
Whether or not customers pays xAI’s premium pricing stays to be seen, significantly given the AI assistant’s tendency to periodically generate politically motivated outputs. These incidents—stemming from deliberate decisions about coaching and system prompts—represenmt basic administration and implementation points that, thus far, no fancy-looking test-taking benchmarks have been in a position to seize.
