grok-build-0.1 API beta prices coding agents at $1/$2 per million tokens

The cost of coding agents is becoming as important as their benchmark score. On May 29, 2026, xAI said grok-build-0.1 is available through the xAI API in public beta. The model is described as the same one that powers the Grok Build CLI, with a focus on agentic coding, and xAI led with a concrete price: $1 per million input tokens and $2 per million output tokens.

“Priced at $1/m input and $2/m output”

That number matters because coding agents burn tokens differently from chatbots. A serious coding task may require repository search, planning, file edits, test output, error analysis, and several rounds of repair. Low unit pricing can make long-running automation more practical, but only if the model is reliable enough to avoid wasteful loops. xAI’s decision to put pricing in the main post suggests it wants developers to evaluate grok-build-0.1 as an operating-cost option, not just a model demo.

The xAI account is the primary channel for Grok model, API, and X ecosystem updates. This post includes the model name, beta status, CLI relationship, and price. Follow-up posts added that the model is also available through OpenRouter and Vercel AI Gateway, and inside tools including Cursor, Hermes Agent, OpenClaw, Kilo Code, and OpenCode. That distribution path is important: coding models gain traction when they appear where developers already run agents.

What to watch next is evidence from real repositories. The relevant tests are not only SWE-style benchmarks but also patch acceptance rate, long-context stability, tool-call discipline, and recovery after failed tests. If the model combines low price with dependable multi-step execution, it could pressure higher-cost coding APIs. If not, the headline price will mostly subsidize failed attempts. source tweet