xAI opens Grok Build 0.1 for API-based coding
xAI has launched `grok-build-0.1` in public beta through its API. The model specializes in agentic coding, delivers more than 100 tokens per second, and costs 1 dollar per million input tokens and 2 dollars per million output tokens.
xAI now offers grok-build-0.1 in public beta through its API, with a model specialized in programming tasks that can run with little human intervention. It is the same model that powers Grok Build.
Its goal is not just to suggest lines of code. It is trained to work as part of an agent: it can help build websites, find and fix bugs, and connect to external tools through MCP, a standard that lets models interact with services and data from other applications.
What Grok Build 0.1 offers
The model generates more than 100 tokens per second. A token is a small unit of text, so this speed translates into fast responses when writing code, reviewing files, or running chained tasks.
The announced price is:
- 1 dollar per million input tokens, meaning the instructions and files the model receives.
- 2 dollars per million output tokens, meaning the responses and code it generates.
Beyond coding, xAI presents it as a fast, affordable option for general-purpose agents and systems that need to call external tools, such as search engines, databases, or internal services.
Who it is for
grok-build-0.1 works best inside an agentic harness: the environment that organizes the instructions, tools, files, and steps a coding agent follows. It is not just the model, but the structure that lets it act on a project.
xAI mentions compatibility with tools such as:
- Grok Build
- Cursor
- Hermes Agent
- OpenClaw
- Kilo Code
- OpenCode
It is also available through OpenRouter and Vercel AI Gateway, two services that let you integrate different AI models without connecting each application directly to a provider's API.
What changes for you
If you develop software, you can try the same model Grok Build uses for tasks such as setting up a website, modifying multiple files, debugging an issue, or coordinating actions with external tools. Its speed and price could be useful in workflows where the agent needs to make many calls in a row.
But there is an important nuance: the model is in public beta, and its performance depends both on the environment running it and on the tools it can access. API availability makes it easier to try, but it does not automatically turn any application into an autonomous programmer.
The next signal worth watching is how it performs outside the environments xAI highlights and how much supervision it still needs in real projects. For now, Grok Build 0.1 expands xAI's push toward agents that can write and modify code directly, with open access for developers to integrate into their own tools.