xAI founder Elon Musk has announced a new version of his language model, Grok 4.20, on the social media platform X. The businessman expressed his hope that this model will take the top spot in the SimpleBench benchmark by the end of August. The announcement followed a user's post highlighting the strong performance of the current Grok 4 version in comparison to the new GPT-5.

Currently, Grok 4 is already outperforming GPT-5 in the ARC-AGI and ARC-AGI-2 tests, while both models show parity in SWE-bench. The improved Grok 4 Heavy version also slightly surpasses GPT-5 Pro in the Humanity's Last Exam test.

It remains unclear exactly what the Grok 4.20 version entails. The official xAI roadmap for this period includes other planned updates, such as Grok 4 Coder and Grok 4 V2, so it is speculated that Musk may have announced one of these projects under a new name.

ORIENT