Elon Musk just made one of the boldest AI predictions of the year: SpaceX could be the best AI company in the world within six months.
On X on Thursday, Musk said he is ‘cautiously optimistic’ that SpaceX will have a model at the level of Anthropic’s Claude Fable and OpenAI’s GPT-6 in two to three months. When a user replied that the rivals would have even better models by then, Musk fired back: if SpaceX’s rate of improvement holds — ‘if our second derivative remains strong’ — SpaceX will reach ‘pole position’ in AI in about six months.
The Grok roadmap
Musk laid out the path in the same thread. Grok 4.7, released this week, sits roughly at the level of Claude Opus 5.0 and placed third in agentic coding rankings behind Anthropic and OpenAI. Next up is Grok 4.8 — a 2.5-trillion-parameter model trained on a brand-new C++ software stack that Musk said was finishing its initial training this week before reinforcement learning begins. Then Grok 4.9, which he expects to be in the class of OpenAI’s Astra and Anthropic’s Fable. And after that: Grok 5, which Musk said might be ‘better than anything.’
Why Musk thinks SpaceX can win
His argument has two parts. First, SpaceXAI is only about three years old, compared with roughly six for Anthropic and ten for OpenAI — so its current position, third place, is actually impressive given its age. Second, and more importantly in his view: compute. ‘Hardware is hard,’ he wrote, but SpaceX has shown it can bring massive amounts of computing power online quickly. In a race where training the best model increasingly comes down to who can deploy the most GPUs fastest, that’s a real card.
Musk also made a characteristically provocative point about intelligence itself: once a model far exceeds the intelligence needed for a class of tasks, additional intelligence is pointless. ‘You don’t need (and it would be cruel to put) Newton-level intelligence in your toaster!’
Where SpaceX stands right now
There’s genuine ground to cover. On Artificial Analysis’ intelligence benchmark, Grok 4.7 scores 46 — behind Claude Opus 5.5’s 58 and GPT-6 Astra’s 53. And there’s a track-record caveat: back in June, Musk promised completely new models trained from scratch every month through the rest of 2026. Grok 4.5, 4.6 and 4.7 have since shipped, but that from-scratch monthly cadence never materialized as promised.
The take
Bold prediction or classic Musk timeline inflation? Either way, it raises the temperature on the three-way AI race at a moment when OpenAI is pausing model training over safety incidents and Google is racing chips into orbit. The next six months should tell us which timeline Musk is actually on.
Comments
Post a Comment