Grok Build launches as xAI joins AI coding-agent race

Elon Musk’s xAI has launched Grok Build, its first AI coding agent, moving the company into a developer-tool market already shaped by Anthropic and other model makers. xAI said in its Grok Build announcement that the early beta lets programmers build software projects from prompts.

The product gives xAI a place in the fight for day-to-day developer work, a stickier prize than occasional chatbot use. Bloomberg reported that Grok Build is the company’s first coding agent and is aimed at rivals including Anthropic, whose coding tools have become a reference point for the category.

Access is still limited.

PCMag said the beta is for subscribers and cited the SuperGrok Heavy subscription at $300 per month. That keeps the launch inside xAI’s paid user base while the company tests whether Grok can handle software-development work rather than general answers alone.

Coding agents are sold on a demanding promise: a model should plan, write, debug and revise code with fewer handoffs between a browser, a terminal and a development environment. xAI is entering that race as teams start judging the tools by how well they fit existing review habits, not by how neatly they answer a prompt.

Mitch Ashley, writing for DevOps.com, put the business case bluntly. “Coding agents are becoming the procurement front where AI labs compete to own the developer workflow,” he wrote.

Ashley said Grok Build can run up to eight parallel AI agents and cited a 70.8 per cent score on SWE-Bench Verified. Those figures give xAI a benchmark claim at launch, although developers and enterprise buyers usually test coding tools against their own repositories before relying on them.

For engineering teams, code generation is only part of the sale. A useful agent also has to keep enough project context to make changes across files, explain the work and avoid creating extra review problems. That is a harder test than producing an answer in a chat window.

A crowded developer race

Anthropic, OpenAI and other AI companies have pushed coding tools because software work is a clear test of whether models can handle long, structured tasks. For xAI, Grok Build extends a product line better known for chatbot features into a market where accuracy, context handling and integration matter more than personality.

Pricing will shape comparisons. DevOps.com cited pricing for grok-code-fast-1 at $0.20 per million input tokens, while PCMag’s report placed the beta inside the higher-priced SuperGrok tier. That gives xAI a subscription signal for heavy users and a model-pricing reference for developers who compare coding agents by cost as well as benchmark performance.

Attention is easier than retention.

A coding assistant keeps users only if it works with a team’s languages, repositories and review habits without adding new errors. Early developer feedback is likely to determine whether Grok Build is treated as a serious tool or as another feature attached to a consumer chatbot.

The launch leaves xAI with the same test facing the rest of the market. In coding agents, the winning products are likely to be judged by time saved on real code, not by a polished demo.