SpaceXAI (formerly xAI) unveils Grok 4.5, built for code and agents

SpaceXAI (formerly xAI) has unveiled Grok 4.5, a new large language model that it calls its most powerful to date. The release is aimed primarily at developers, engineers and others who work on technical tasks.

Who is Grok 4.5 for?

According to the company, the model was built with code writing, agentic workflows and everyday intellectual work in mind. That sets Grok 4.5 apart from general-purpose models, which try to handle creative writing, everyday questions and programming equally well at the same time, often sacrificing quality in each area for the sake of versatility.

• Developer: SpaceXAI, the company previously operating as xAI
• The model is called Grok 4.5 and is positioned as the most powerful in the company's lineup to date
• Training was carried out jointly with the team behind the Cursor development environment
• According to SpaceXAI, the model spends tokens about twice as efficiently as competitors
• The model's main focus is code, agentic workflows and technical work

Why the Cursor partnership matters

Cursor is one of the most prominent development environments with a built-in AI assistant, used by programmers around the world. Training the model jointly with the Cursor team means Grok 4.5 was tuned for real programmer workflows from the start, rather than adapted to them after the release of a general model. For a developer, this approach usually means a more accurate grasp of code context, fewer erroneous suggestions and more natural behavior inside the development environment itself, where the model was also tested during training.

What halving token use means

SpaceXAI's claim that Grok 4.5 uses tokens about twice as efficiently as competitors bears directly on agentic workflows, meaning situations where the model carries out a chain of actions on its own: it reads files, calls tools, edits code, checks the result and repeats the loop as needed. In such workflows the model can make dozens or hundreds of calls in a row, and token costs quickly add up to a noticeable sum for a team building a product on top of the model. If efficiency really is twice that of competitors, it directly lowers the cost of running agents built on Grok 4.5.

Where the market for development models is heading

A couple of years ago, most labs released general-purpose models and only later adapted them for code through fine-tuning and dedicated modes. SpaceXAI took a different path: building Grok 4.5 around developer tasks and agentic workflows from the outset, rather than retrofitting a general-purpose model. The partnership with Cursor during training is part of the same strategy: instead of releasing a model "in a vacuum" and waiting for the programming community to adapt it to its tools, the company embeds the model in a workflow developers already use.

This approach also reflects a shift in the industry's monetization model: agentic workflows, where AI carries out long chains of actions on its own, require not a one-off answer to a question but stable, predictable and cheap operation over hundreds of steps. That is why SpaceXAI emphasizes token efficiency on a par with the model's capabilities: for those building products on Grok 4.5, the two parameters are closely linked.

What this means

SpaceXAI's bet on developers, code and agentic workflows shows where competition between AI labs is shifting: the winner is not only the model that is "smarter" in a general sense, but the one that is cheaper and faster in programmers' real working cycles. Working with the Cursor team on training is a signal that major AI labs are increasingly building partnerships with the makers of tools that millions of programmers already use, instead of competing with them directly for the user.