Grok Code Fast 1 is an economical, fast coding assistant from xAI designed for agentic coding workflows. Built with a new model architecture and a programming-focused training corpus, it’s refined with real-world datasets and partner feedback, enabling it to use common developer tools (grep, terminal, editing) and feel native in IDEs. For a limited time it’s available for free with exclusive launch partners, including GitHub Copilot, Cursor, Cline, Roo Code, Kilo Code, opencode, and Windsurf. Pricing is $0.20 per 1M input tokens, $1.50 per 1M output tokens, and $0.02 per 1M cached input tokens. The team highlights blazing speed, aided by prompt caching with high cache-hit rates, so the model often calls dozens of tools before you finish reading. Grok Code Fast 1 is versatile across TypeScript, Python, Java, Rust, C++, and Go, capable of tasks from zero-to-one projects and codebase questions to surgical bug fixes. A battle-sim demo and a developer quote illustrate its speed in practice. A “Grok Code for everyone” section lists partner logos and ongoing free access, plus resources like a Prompt Engineering Guide and a model card. In testing, the model scored 70.8% on SWE-Bench-Verified tasks in an internal harness and is guided by human evaluations for usability and satisfaction. Looking ahead, rapid iterations and a multimodal variant with parallel tool calls and extended context length are in training, with a new model card to follow; the stealth release was code-named sonic.