xAI have introduced grok-code-fast-1, "a speedy and economical reasoning model that excels at agentic coding" with free access through launch partners for a limited time.
xAI "built grok-code-fast-1 from scratch, starting with a brand-new model architecture." The development process involved carefully assembling "a pre-training corpus rich with programming-related content" and curating "high-quality datasets that reflect real-world pull requests and coding tasks" for post-training.
On evaluation benchmarks, "grok-code-fast-1 scored 70.8% using our own internal harness" on "the full subset of SWE-Bench-Verified." xAI noted that "while benchmarks like SWE-Bench provide valuable insights, we've found they don't fully reflect the nuances of real-world software engineering, particularly the end-user experience in agentic coding workflows."
The model is "exceptionally versatile across the full software development stack and is particularly adept at TypeScript, Python, Java, Rust, C++, and Go." It "can complete common programming tasks with minimal oversight, ranging from building zero-to-one projects and providing insightful answers to codebase questions to performing surgical bug fixes."
xAI designed the model to be "widely accessible" with pricing at:
- $0.20 per million input tokens
- $1.50 per million output tokens
- $0.02 per million cached input tokens
xAI "teamed up with select launch partners to offer grok-code-fast-1 for free for a limited time, including GitHub Copilot, Cursor, Cline, Roo Code, Kilo Code, opencode, and Windsurf."
Mario Rodriguez, Chief Product Officer at GitHub, stated: "In early testing, Grok Code Fast has shown both its speed and quality in agentic coding tasks. 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." Additionally, "a new variant that supports multimodal inputs, parallel tool calling, and extended context length is already in training."