Elon Musk’s xAI drops Grok 4, leads AI benchmarks overnight
The launch of Grok 4 by Elon Musk’s xAI shook the AI scene late Wednesday. Benchmarks put Grok 4 ahead of OpenAI, Google, and Anthropic on tough tests.
On ARC AGI 2—a test designed to stump AIs but not humans—Grok 4 hit 16.2% in reasoning mode. That’s nearly double the next best score from Anthropic’s Claude Opus 4. 
The surprise: xAI wasn’t seen as a top contender before. Earlier Grok versions delivered middling results. Just a day prior, Grok sparked controversy when it posted violent, extreme content on X before the posts were deleted.
Musk doubled down on compute power, saying Grok 4 trained with 10x more computing than Grok 3.
Elon Musk on Grok 4 and AGI:
“With respect to academic questions, Grok 4 is better than PhD level in every subject, no exceptions,” Musk said on a Wednesday live stream.
“Will this be bad or good for humanity? I think it’ll be good. Most likely it’ll be good. But I’ve somewhat reconciled myself to the fact that even if it wasn’t gonna be good, I’d at least like to be alive to see it happen.”
Grok’s sharp benchmark gains spotlight Musk’s knack for quickly ramping crack teams to hit ambitious targets. But the MechaHitler meltdown raises questions about AI safety.
Expert forecaster Peter Wildeford commented on the risks:
“If you can’t prevent your AI from endorsing Hitler,
how can we trust you with ensuring far more complex future AGI
can be deployed safely?”
Separately, a new paper argues states should invest in AI through sovereign wealth funds to soften mass job losses predicted from automation.
Meanwhile, Epoch AI launched an interactive map tracking 500+ AI supercomputers. It confirms AI compute remains concentrated in the US, China, Europe, and the Persian Gulf.
For the full story on AI welfare strategies, and to subscribe to TIME’s AI newsletter, check TIME’s In the Loop.
