SpaceXAI shipped Grok 4.5, the company's first model developed through its partnership with Cursor and its most capable release since rebranding from xAI.

The model was trained on tens of thousands of NVIDIA GB300 GPUs using datasets focused on coding, science, engineering and mathematics. SpaceXAI said Grok 4.5 was built specifically to excel at coding, agentic tasks and knowledge work.

The company claims the model outperforms competing systems at real engineering tasks and can create functional applications with minimal prompting. In one demonstration, Grok 4.5 generated an interactive solar system simulation from a single prompt.

Grok 4.5 costs $2 per million input tokens and $6 per million output tokens. That undercuts OpenAI GPT-5.6's most powerful Sol variant, which costs $5 per million input tokens and $30 per million output tokens, though OpenAI's cheapest Luna variant matches Grok 4.5's output pricing at $6 per million tokens.

The model now powers SpaceXAI's terminal-based coding agent Grok Build, which handles coding tasks as well as Excel, PowerPoint and Word automation. Grok 4.5 is also available through the SpaceXAI console and integrated into all Cursor plans.

"We've partnered with SpaceXAI to train Grok 4.5," Cursor said on X. "It's our most powerful model yet and the first we've built for more than software engineering."

SpaceXAI and Cursor announced their partnership in April, with terms that could see SpaceXAI either investing $10 billion into Cursor or acquiring the company outright for $60 billion later this year. The companies have not disclosed which path they will pursue.

The model remains unavailable in the European Union, where SpaceXAI expects to launch access in mid-July.