Elon Musk's SpacexAI Unveils Grok 4.5, Faster Coding, Smarter Reasoning & Lower Costs

According to AI coding platform Cursor, Grok 4.5 is capable of solving difficult tasks that involve multiple repositories, hundreds of tools and lengthy workflows. The company said the model can creatively use external tools to complete tasks that developers typically perform on a computer.

Grok 4.5 is a Mixture-of-Experts (MoE) model jointly trained with Cursor. The training process reportedly included trillions of tokens collected from software repositories and developer interactions, allowing the model to understand better how programmers write code, debug software, and interact with development tools.

SpacexAI said the model was trained on tens of thousands of NVIDIA GB300 GPUs, using new large-scale training and stability techniques.

The company also invested heavily in filtering and curating datasets, removing duplicate content and selecting high-quality, domain-specific training data to improve overall accuracy.

Built for Real-World Software Development

The company says Grok 4.5 is designed specifically for engineering workloads rather than simple chatbot conversations.

It can reportedly:

• Handle large enterprise codebases.

• Work across multiple repositories simultaneously.

• Solve advanced Rust, C and C++ programming challenges.

• Build complete applications from a single prompt.

• Use external development tools during long-running tasks.

• Maintain context across lengthy coding sessions.

According to xAI, the model delivers Opus-class performance while offering faster responses, improved token efficiency and lower operating costs.

What's New in Grok 4.5?

Compared to previous Grok models, Grok 4.5 introduces several major improvements:

• More advanced reasoning for complex technical problems.

• Improved coding capabilities across multiple programming languages.

• Better support for long-running, multi-step workflows.

• Ability to work with hundreds of developer tools and software environments.

• Improved context retention for large projects.

• Higher-quality training data with extensive filtering and deduplication.

• Faster inference and lower token costs.

• Optimised for production-level software engineering instead of only conversational AI.

How It Differs from Previous Grok Versions

Earlier Grok models, including Grok 3 and Grok 4, focused heavily on conversational AI, reasoning, web search integration and general-purpose assistance.

While they already offered coding support, they were primarily designed as chatbot-style AI assistants integrated into X.

Grok 4.5 shifts that focus significantly towards developer productivity.

Unlike its predecessors, the latest model is trained specifically to:

• Understand large software projects.

• Navigate complex codebases.

• Complete long-duration engineering tasks.

• Interact with multiple software tools during execution.

• Generate production-ready applications from prompts.

The company says these enhancements make Grok 4.5 one of its strongest coding-focused AI models so far.

Availability and Pricing

Grok 4.5 is available starting today through:

• Cursor Desktop

• Cursor Web

• iOS

• CLI

• Cursor SDK

The pricing includes:

• Base model: $2 per million input tokens and $6 per million output tokens.

• Fast variant: $4 per million input tokens and $18 per million output tokens.

Cursor said individual and team subscriptions include significant access to the model through its first-party model pool. To mark the launch, the company is also doubling user limits during the first week.

EU Launch Coming Soon

At launch, Grok 4.5 is not available in the European Union. Cursor says support for EU users is expected later this month.

With its stronger focus on enterprise software development, long-context reasoning and production-scale coding, Grok 4.5 represents xAI's biggest step yet toward competing with leading AI coding models in the developer ecosystem.