SpaceXAI launched Grok 4.5 on July 8, 2026, its newest AI model and the first one created with help from a company called Cursor, which writes software Axios. The model was trained on massive computers in SpaceXAI's own data centers, using information focused on coding, science, engineering, and math x.ai.

SpaceXAI says Grok 4.5 is designed to help people write code, handle tasks that require multiple steps (like having an AI assistant break down a project and complete it piece by piece), and answer general questions. The company is calling it its "smartest model" yet.

How Much Does It Cost?

Grok 4.5 costs $2 per million input tokens (the words you type in) and $6 per million output tokens (the words the AI generates back). Think of tokens as roughly equivalent to words—a few characters grouped together Engadget.

For comparison, OpenAI's GPT-5.6 Sol costs $5 for input and $30 for output. Another model called Luna costs $1 for input and $6 for output. This makes Grok 4.5 roughly in the middle on input cost but much cheaper on output.

That output price difference matters more than it might seem. When an AI is asked to handle multi-step projects—like writing code where it has to look things up, adjust its approach, and try again—those workflows generate a lot of output tokens. If Grok 4.5 works as well as competitors but costs a fraction as much on those jobs, companies might choose it simply for the money saved.

Where You'll Use It

Grok 4.5 is now the default model in Grok Build, SpaceXAI's tool for coding projects. You can also access it through Cursor, a software writing assistant, and it works across all of Cursor's pricing tiers.

SpaceXAI says the model also handles tasks in Excel, PowerPoint, and Word—not just code. Cursor announced the partnership on X (formerly Twitter) and published details on its website Cursor.

The Deeper Business Deal

In April, SpaceXAI and Cursor signed a partnership that goes beyond simply using each other's technology. The agreement includes an option for SpaceXAI to invest $10 billion in Cursor, or to acquire the company outright for $60 billion later this year Engadget.

This is important because it means Grok 4.5 was not just licensed to Cursor as a separate product. Instead, Cursor's engineers actually helped train the model using their knowledge of how developers really use AI assistants—which suggestions they accept, which ones they rewrite, where the AI gets things wrong. This kind of deep collaboration is different from the standard practice of one company simply buying access to another company's AI through an API (a technical connection that lets two software products talk to each other).

The shift also matters for what Cursor is betting on going forward. The company built its reputation by staying open to multiple AI models, letting customers choose which AI they preferred. Now Cursor's future is more closely tied to SpaceXAI's direction, which is a significant change in strategy.

The rebranding from xAI to SpaceXAI is worth noting too. It signals that SpaceXAI is now more directly part of Elon Musk's larger SpaceX organization than before, though the companies haven't provided details on exactly how that integration works Reuters.

What SpaceXAI Claims—and What We Don't Know Yet

SpaceXAI says Grok 4.5 outperforms other top-tier models at real engineering tasks and can write working applications from brief instructions. Those are claims made by the company itself, not verified by independent testing. As of now, neither xAI nor Cursor has published results from outside evaluators x.ai.

That matters. In this author's view, the strongest test will come over the next few weeks, when developers and companies actually use Grok 4.5 on real projects and report back on whether the model's performance matches its promises. Vendor benchmarks are useful starting points, but they do not replace real-world evidence.

The European Delay

Grok 4.5 is not yet available in the European Union. SpaceXAI says it expects to bring the model to EU customers by mid-July, roughly a week to ten days after the US launch. Neither company specified why, though the gap likely reflects regulatory review related to AI rules in Europe.

What This Means for People Looking for Coding Help

If you work in software development or at a company evaluating coding assistants, Grok 4.5 is now a credible option alongside OpenAI and Anthropic's Claude models. The combination of lower pricing (especially on output tokens), built-in support in Cursor, and an expanding toolkit for office tasks makes it a genuine alternative on paper.

Whether it works equally well in practice is the real question. That answer will emerge as developers put it to work over the coming weeks.