AI BriefingGrok Code Fast 1 Released
Key point
xAI has released Grok Code Fast 1, a fast and low-cost reasoning model optimized for agentic coding workflows.
Details
xAI has released grok-code-fast-1, a fast and low-cost reasoning model optimized for agentic coding workflows. Recently stealth-launched under the codename 'sonic,' this model is designed to minimize latency in agent-based coding environments that involve frequent iterative reasoning and tool calls.
It was built from scratch on a new model architecture and pretrained on a dataset rich in programming-related content. In particular, it has mastered the use of developer tools such as grep, terminal, and file editing, making it immediately usable in IDE environments.
Key Features and Performance:
- Overwhelming Speed: Delivers extremely fast response times through innovative reasoning technology, achieving a cache hit rate of over 90% via prompt caching optimization.
- Broad Language Support: Proficient in major languages including TypeScript, Python, Java, Rust, C++, and Go.
- Excellent Benchmarks: Scored 70.8% on the full SWE-Bench-Verified subset.
It is currently offered free for a limited time through major partners such as GitHub Copilot, Cursor, Cline, and Windsurf. Pricing is highly economical at $0.20 per million input tokens, $1.50 per million output tokens, and $0.02 per million cached input tokens.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.