Today we are releasing Grok 4.5 together with SpaceXAI, our most intelligent model and the first we’ve built for more than software engineering. It’s available today in Cursor!
Grok 4.5 is built for difficult, long-running tasks that require creative use of tools to solve problems, whether that’s software engineering, data science, finance, legal work, or anything else you do on a computer.
What’s new:
-
Our first model trained with SpaceXAI: Grok 4.5 was trained jointly with SpaceXAI on a new pretrained mixture-of-experts model, then continued on trillions of tokens of Cursor data. Where Composer specialized in coding, we kept Grok 4.5’s training mix deliberately broader so it’s proficient far beyond software.
-
RL on difficult problems: We used reinforcement learning on problems designed to be so hard that even frontier models fail at them, built with a new distributed agent system. Some environments would have taken teams of hundreds of engineers months to create.
-
Effort levels: High, medium, and low, so you can match compute to how hard the task is.
Grok 4.5 is live across desktop, web, iOS, the CLI, and our SDK, with significant included usage on individual and team plans.
Grok 4.5 will become available in the EU in the coming weeks.
And Composer isn’t going anywhere: it remains our fast, cost-efficient model for everyday coding, with a new version coming soon. 
We’d love for you to share your first impressions.