Grok Code Fast 1 revolutionizes proxy coding

xAI launches Grok Code Fast 1 for proxy coding
xAI has announced the release of its new Grok Code Fast 1, a fast and efficient inference model specifically designed to optimize proxy coding performance in development environments. Focused on speed, precision, and versatility, the model responds to the everyday needs of developers who struggle with slow tool calls and for loops in existing models, which can hinder the coding workflow.

Exceptional performance and speed
The model records 190 symbols per second (TPS), outperforming leading models such as Claude Sonnet 4, Gemini 2.5 Pro, GPT-5, Qwen3-Coder, and Grok 4, which range from 150 to 175 TPS. xAI relies on heuristics and supercomputing techniques to boost the speed of tool invocation, while optimized caching that achieves response rates of over 90% when used with launch partners, providing a seamless and immediate experience for developers.

Versatility and flexibility
The model supports TypeScript, Python, Java, Rust, C++, and Go programming languages, and can perform multiple tasks including building projects from scratch, accurately answering database queries, and fixing bugs efficiently with minimal supervision. This flexibility makes it suitable for developers of all experience levels, both beginners and professionals.

Design and architecture
Grok Code Fast 1 was developed from scratch using a new modular architecture, and xAI collected high-quality datasets reflecting real-life coding tasks before and after the training phase. The company's teams also collaborated with launch partners such as GitHub Copilot, Cursor, Cline, Roo Code, Kilo Code, opencode, and Windsurf to ensure smooth performance in different IDE environments, with excellent ability to interact with popular tools such as grep and terminal.

Pricing and economy
The model is designed to be economical and scalable, with a cost per use of $0.20 per million input codes, $1.50 per million output codes, and $0.02 per million cached input codes. This balance of high performance and low cost makes the model a practical choice for processing programming tasks quickly and efficiently.

Partnerships and Free Launch
The model is available for free for a limited time to exclusive launch partners such as GitHub Copilot, Cursor, and Cline, to further empower developers with powerful tools. GitHub's Mario Rodriguez expressed his enthusiasm saying: "Grok Code Fast has proven its speed and quality in proxy coding tasks, and empowering developers with powerful tools is a key part of our mission."

Future Development and Improvements
xAI plans to release rapid updates that include support for multimodal input, calling tools in parallel, and lengthening the task context to improve model understanding. The company has also released a Rapid Engineering Guide to guide developers and make the most of the model's capabilities.

Assessing Performance and Potential Gaps
The company relied on SWE-Bench and human evaluations to ensure performance balance, accuracy, and quality of the final experience. The model received a score of 70.8%, reflecting its efficiency in real-world, everyday tasks. However, the report does not yet reveal precise details about the model's performance in large multi-team software project situations or how the model interacts with unmentioned third-party tools, a potential gap in practical evaluation.