Scoop: SpaceXAI launches new model, Grok 4.5
TL;DR
Axios reports that SpaceXAI is releasing Grok 4.5, its strongest model so far and its first major launch since going public and acquiring Cursor. The pitch is aimed at coding, engineering and agentic knowledge work, not mainly at consumer chatbot use. Musk frames Grok 4.5 as Opus-class: faster, more token-efficient and cheaper. The benchmark chart still comes from the company’s own launch material. Grok 4.5 is available in Grok Build, Cursor on all plans and the SpaceXAI console. It is not available in the EU yet.
Nauti's Take
This is a strong launch claim, but not yet clean proof. SpaceXAI is selling Grok 4.5 aggressively as a coding and agent model, while the main performance story still comes from its own benchmark framing.
Cursor is the part to watch: if the model is deeply embedded there, developers will feel the difference faster than any leaderboard can explain. EU users are stuck watching from the outside for now.
Briefingshow
Grok 4.5 shows where model competition is moving: away from generic chatbot comparisons and toward real work surfaces such as IDEs, build environments and agent stacks. For teams, the key question is less which model sounds smartest in a demo chat. It is which model can handle code, context, cost and tool access inside the workflow without creating new friction.
Sources