Cursor and SpaceXAI released Grok 4.5 — the company’s first model beyond programming

Cursor, in partnership with SpaceXAI, released Grok 4.5 on July 9, 2026 — a new flagship model with a mixture-of-experts architecture that, for the first time in the company's lineup, is oriented not only toward programming, but also toward data science, finance, legal work, and other knowledge work tasks.

What Grok 4.5 Can Do

The new model was developed as a universal tool for complex, long-running tasks. Unlike its predecessor Composer 2.5, which was specialized exclusively for code, Grok 4.5 was trained on an expanded dataset — STEM tasks, scientific papers, and materials from various professional fields.

• Architecture: mixture-of-experts, joint development by Cursor and SpaceXAI
• Training: trillions of tokens of data on user interactions with code repositories
• Areas of application: programming, data analytics, finance, legal work
• Base price: $2 per 1 million input tokens and $6 per 1 million output tokens
• Fast variant: $4 per 1 million input tokens and $18 per 1 million output tokens
• Availability: desktop, web, iOS, CLI, and SDK

How Cursor Trained the Smartest Model

The key element of training is reinforcement learning on tasks in realistic environments that encompass both software development and broader intellectual work. The environments were specifically designed to be challenging enough that even state-of-the-art models would struggle with them.

To scale the process, Cursor developed a distributed agent system: engineers set a task and verification criteria, and groups of agents automatically create, test, and refine training environments. According to the company, some of these would have required teams of hundreds of people several months of manual work.

Cursor also openly reported an important caveat: Grok 4.5 showed an advantage on the internal CursorBench benchmark because an early snapshot of Cursor's codebase accidentally made it into the training data. The exact scale of the effect is unknown. The company has already removed this data for future models and is working on updating the benchmark.

Pricing and How to Get Access

Grok 4.5 is already available in all versions of Cursor — desktop, web, iOS, CLI, and SDK. Subscribers to individual and team plans receive a significant volume of model usage within their pool, and during the first week the company doubles this limit. Composer 2.5 remains in the offering, and Cursor plans to release new models of this class in the future.

Along with the model's release, the company added new security mechanisms that leverage Grok 4.5's cybersecurity capabilities — their specifics are not disclosed, but the fact of their inclusion indicates that the model is assessed as a potentially powerful tool in this area as well.

What This Means

Cursor is betting on a shift from narrowly specialized coding assistants to universal agents for knowledge work. Grok 4.5 is the company's first attempt to move beyond the developer niche and offer a tool that works equally well with code, legal documents, and financial data.

Frequently Asked Questions

How is Grok 4.5 different from Composer 2.5?

Composer 2.5 was specialized solely for programming. Grok 4.5 was trained on a broader dataset — STEM tasks, scientific papers, knowledge work — and is designed for data analytics, finance, and legal work, not just development. Cursor plans to develop both model classes in parallel.

How much does it cost to use Grok 4.5 via API?

The base version costs $2 per 1 million input tokens and $6 per 1 million output tokens; the fast variant costs $4 per 1 million input tokens and $18 per 1 million output tokens. Subscribers to Cursor's individual and team plans receive a significant volume of usage for free — within their primary models pool.