xAI ships grok-build-0.1 coding API and pledges Grok 4.2 open-source by year-end

xAI on May 29 made grok-build-0.1 available in public beta through the xAI API at $1 per million input tokens and $2 per million output tokens — aggressive pricing for an agentic coding model. It's the same model powering the Grok Build CLI, xAI's coding agent. The launch was paired with a dedicated public page tracking daily build version updates, an unusual transparency move for a frontier lab and a clear positioning play against Anthropic Claude Code and Microsoft GitHub Copilot.

The bigger strategic news came earlier in the week: Elon Musk confirmed on May 25 that training for Grok V9 Medium, a 1.5 trillion parameter foundational model, has completed, with public release targeted mid-June 2026. Musk also pledged to open-source the 0.5T-parameter Grok 4.2 base model by end of 2026 — a credible answer to DeepSeek's price-cutting and the broader open-weight momentum from China. Grok 5 is training on Colossus 2, a 1-gigawatt cluster with 500,000 GPUs that xAI says was built in 122 days and can train seven models simultaneously, with realistic release estimates in Q3–Q4 2026.

Hacker News reaction was mixed. The Grok 4.2 open-source pledge was seen as credible and timely against DeepSeek and local-LLM momentum, but Musk's parallel hype that Grok 5 will be 'indistinguishable from AGI' undercut trust in the timeline — a familiar pattern. xAI's daily build updates and Grok Build pricing got more developer respect than the AGI framing.

Competitive context: at $1/$2 per million tokens, grok-build-0.1 is roughly half the price of equivalent Anthropic and OpenAI coding endpoints, but trails Claude Code and GitHub Copilot on developer mindshare. Watch next: Grok V9 Medium's actual benchmark performance in mid-June, whether the 4.2 open-source release actually ships in 2026 (Musk timelines slip), and whether Colossus 2's seven-models-simultaneously claim holds up.