xAI Unveils Grok Code Fast 1, Optimised for Agentic Coding

xAI has launched Grok Code Fast 1, a new reasoning model designed for agentic coding workflows, offering faster tool use and lower costs compared to existing options. The model is available for free for a limited time through partners including GitHub Copilot, Cursor, Cline, Roo Code, KiloCode, opencode and Windsurf.

Grok Code Fast 1 is now generally available through the xAI API.

The company stated that Grok Code Fast 1 was developed from scratch with a new architecture and trained on programming-rich datasets, followed by post-training on real-world coding tasks, including pull requests and bug fixes. The model has been integrated into agentic coding platforms where it supports tools such as grep, terminal and file editing.

“In early testing, Grok Code Fast 1 has shown both its speed and quality in agentic coding tasks,” Mario Rodriguez, chief product officer at GitHub, said. “Empowering developers with powerful tools is a core part of our mission at GitHub Copilot, and this is a compelling new option for our developers.”

According to xAI, the model processes at 190 tokens per second, significantly ahead of rivals like GPT-5, Gemini 2.5 Pro and Claude Sonnet 4. It is priced at $0.20 per million input tokens, $1.50 per million output tokens and $0.02 per million cached input tokens.

The company also reported benchmark results, with Grok Code Fast 1 scoring 70.8% on the SWE-bench Verified subset, measured using its internal evaluation harness. 

However, xAI emphasised that human assessments during real-world workflows played a larger role in refining the model.

Danny Limanseta, an early tester, said on X that the system was so quick he “had to change up how I work in Cursor.”

xAI noted that the model is versatile across various programming languages, including TypeScript, Python, Java, Rust, C++, and Go, and can handle tasks ranging from project creation to bug fixes.

The release follows a stealth phase where the model, internally codenamed ‘Sonic,’ was tested with developers before its wider rollout. xAI said updates will arrive frequently, with a multimodal variant already in training that will support parallel tool calling and extended context lengths.