Elon Musk thinks Grok 5 might actually hit AGI

Elon Musk used to be the loud voice warning that unchecked artificial intelligence could end humanity. Now, he is betting his own startup is on the verge of cracking artificial general intelligence first.

Musk posted on X that he now believes xAI has a genuine shot at reaching AGI with its upcoming Grok 5 model, openly confessing that he “never thought that before.”

It is one of the boldest claims yet in the current AI race, directly attaching the elusive AGI milestone to a specific upcoming version.

The benchmark run behind the claim

Musk’s sudden optimism seems fueled by recent test performances. He chimed in after user @XFreeze pointed out that Grok 4 had broken records across AGI benchmarks, particularly in complex problem-solving and open program synthesis.

According to those reports, Grok 4 outpaced rival systems, leading Musk’s supporters to claim it is currently closer to AGI than any other system on the market.

Musk has never been shy about hyping his chatbot. He previously labeled Grok the “smartest AI in the world” and claimed it was “better than PhD level” across every academic discipline.

Interestingly, when Grok 4 launched back in July, it came with a distinct twist: the model was tuned to lean on Musk’s own posts on X to help guide its responses.

A dramatic shift and a messy record

Musk’s current sprint toward superintelligence stands in sharp contrast to where he stood two years ago.

Roughly eight months before Grok officially launched in November 2023, Musk was the marquee signatory on an open letter demanding an immediate pause on advanced AI model training. Co-signed by Apple co-founder Steve Wozniak and researchers from Google DeepMind, the letter warned that human-competitive AI could pose profound existential risks.

Instead of slowing down, Musk built xAI and rolled out Grok as an unapologetic, “anti-woke” competitor to OpenAI and ChatGPT.

By giving X users free access and rolling out standalone web and mobile apps, xAI quickly pulled in tens of millions of users. But the platform has repeatedly courted controversy along the way.

Back in July, the US Anti-Defamation League blasted Grok as “irresponsible, dangerous and antisemitic” after the bot praised Adolf Hitler in some interactions.

More recently, Grok drew flak for hallucinating false accusations against London’s Metropolitan Police over protest footage.

Why it matters

For everyday users and builders, take the AGI label with a healthy grain of salt. Grok 4 has certainly made strides in code generation and problem-solving, but an AI that still trips over basic factual reporting is not ready to replace human cognition across the board.

For the industry, Musk is turning up the competitive heat. OpenAI, Meta, and Google are all chasing their own versions of human-level intelligence, and Musk’s public declaration forces competitors to justify their timelines to investors and talent.

We think Musk’s move is classic PR showmanship: anchoring xAI as the frontrunner in the AGI narrative without offering technical proof. The real hazard is obvious. When you slap the AGI label onto a model that is still ironing out severe bias and hallucination bugs, you risk trivializing the concept and burning public trust when Grok 5 inevitably behaves like just another software tool.