Grok Code Fast 1 Lands in Cline: Free Launch Week

TL;DR

• xAI's Grok Code Fast 1 (the "Sonic" stealth model) is now public and integrated into Cline
• Built specifically for agentic coding workflows with 90%+ cache hit rates and sub-second response times
• Free during launch week; then $0.20/M input tokens, $1.50/M output tokens, $0.02/M cached tokens
• Update Cline and select grok-code-fast-1 from model options — no config needed

What Dropped

xAI unveiled Grok Code Fast 1, a model architected from scratch for agentic development. It's been quietly available as "Sonic" in Cline for the past week; now it's official. The model is free during launch week, then moves to xAI's standard pricing.

The Dev Angle

This isn't a general-purpose model repurposed for coding. xAI built Grok Code Fast specifically for iterative development loops — the kind Cline runs constantly. The architecture prioritizes speed and tool use over raw capability, which means faster feedback cycles and less time waiting for model responses.

The numbers matter: 90%+ cache hit rates in typical Cline workflows mean subsequent requests feel nearly instantaneous. The model excels across TypeScript, Python, Java, Rust, C++, and Go. It handles everything from greenfield projects to surgical bug fixes, and it understands complex codebases well enough to maintain consistency across long development sessions.

Integration is frictionless. Select grok-code-fast-1 from Cline's model dropdown and start coding. It works with Plan/Act modes, MCP tools, and all existing Cline features — no special configuration required. If you've been testing the Sonic stealth model, you already know what to expect.

Should You Care?

If you're using Cline for iterative development, this is worth testing during the free week. The speed advantage is real — you'll notice it immediately in how quickly the model calls tools and iterates on solutions. The workflow shift from "craft one massive prompt" to "give focused tasks and iterate rapidly" actually changes how you work.

The pricing is competitive for what you get. At $0.20/M input tokens and $1.50/M output tokens, it's not the cheapest option, but the cache efficiency (90%+ hit rates) and speed mean you'll spend less time waiting and fewer tokens overall on iterative tasks. If you're already paying for Cline integration with other models, the cost difference is marginal.

Skip this if you're doing one-shot code generation or working on projects where latency doesn't matter. Grok Code Fast is optimized for the tight feedback loops that define agentic development — if that's not your workflow, the speed advantage disappears.

Update Cline to the latest version and select grok-code-fast-1 from your model options. Share your results on Reddit or Discord.

Source: Cline