Elon Musk has put a timeline on xAI's next frontier model. In a July 18 exchange on X, Musk said the company's new 2-trillion-parameter system — the successor to the roughly 1.5-trillion-parameter model behind Grok 4.5 — will "finish initial training next week."
What Musk claimed
He described the model as "better than our 1.5T in every way" and said it "might be able to exceed Kimi" — a hedged nod to Moonshot AI's 2.8T Kimi K3 rather than a firm benchmark claim. Crucially, he added that speed and token efficiency should stay "close to our 1.5T (aka Grok 4.5)," signaling the jump is not a brute-force scale-up that trades cost for capability.
The name
Pressed on whether it would be called Grok 4.6, Musk replied "Yeah, Grok 4.6." That one-word confirmation is as official as the naming gets so far; xAI, which now operates under the combined SpaceXAI branding, has not issued a formal product announcement.
When it ships
The July 18 post addressed only training. A separate, earlier statement from Musk indicated the 2T model would be available to customers "next month" — pointing to an August 2026 window — but xAI has published no confirmed release date, and initial training finishing is a milestone well short of general availability.
The scale race
A roughly 33% parameter increase lands xAI in the same weight class as the largest Chinese open models, and continues a year in which Musk has teased an aggressive roadmap stretching toward a 10-trillion-parameter system. The efficiency promise is the part to watch: frontier labs are increasingly judged not on raw size but on cost per useful token, and Musk is explicitly claiming the 2T model will not blow up inference costs.