Grok Build 0.1 launched in public beta on approximately May 29. That launch brief covered what developers could test and what remained unverified. The June 5 update added something genuinely useful for teams running parallel development tasks: native Git worktrees support.
Git worktrees is an established Git feature, not xAI’s invention, that lets developers work on multiple branches simultaneously in separate working directories on a single repository. xAI’s implementation in Grok Build means the AI subagents handling parallel coding tasks can now operate in isolated directories, preventing the conflicting rewrites on the main branch that agentic coding tools produce when multiple agents share workspace. For teams running multi-thread agentic development workflows, this addresses a real coordination problem. The xAI blog source for this announcement is currently inaccessible, the feature description is vendor-stated.
The bigger forward-looking signal in the June 5 update: according to xAI, pre-training for Grok V9-Medium is complete. The model is reportedly in Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) phases, with a mid-June 2026 public beta as the stated target. These details come from xAI’s announcement and statements attributed to Elon Musk, they’re from the same entity, not independently corroborated. The source is unavailable. Use this timeline as xAI’s stated intent, not a confirmed delivery schedule.
On the sizing claim: xAI describes V9-Medium’s scale as approximately 1.5 trillion, but the original package used “tokens” and “parameters” in adjacent clauses while deriving a 3x size comparison to Grok v8-small. Training tokens and model parameters are different metrics. A Wire re-research request is open to clarify whether the 1.5T figure refers to training data volume or parameter count. The specific comparative sizing claim doesn’t appear in this brief until that’s resolved. What’s safe to say: xAI describes V9-Medium as significantly larger than its current production model.
Unanswered Questions
- Is the 1.5T figure training token volume or parameter count, and what does the correct interpretation mean for comparing V9-Medium to Grok v8-small?
- When will xAI publish a DPA for Grok Build, and what data processing terms will it include for enterprise users?
- Does the mid-June beta target include EU-market availability, and if so, under what compliance framework?
Pricing, per xAI’s stated figures as of June 5: $1.00 per million input tokens and $2.00 per million output tokens. SuperGrok Heavy subscription required, currently listed at a reported $99/month early-adopter rate from a $300/month list price. Verify current pricing at checkout, vendor pricing can change.
European developers should note this. Grok Build involves code and potentially proprietary data passing through xAI’s infrastructure. As of June 7, xAI has not published a standard Data Processing Agreement (DPA) for Grok Build. That gap creates friction under GDPR Article 28 for any European enterprise user, not a confirmed violation, but a compliance consideration that should sit in your vendor assessment before committing. The EU AI Act also carries transparency obligations relevant to AI-assisted coding tools. Document the DPA gap; follow up with xAI before signing enterprise agreements.
Platform note: Grok Build is currently macOS and Linux only. Windows users need WSL2.
Warning
xAI has not published a Data Processing Agreement (DPA) for Grok Build as of June 7, 2026. European enterprise teams whose developers use Grok Build are passing code and potentially proprietary data through xAI infrastructure without standard GDPR Article 28 contractual protections in place. This is not a confirmed violation, it is an unaddressed compliance gap that requires vendor engagement before enterprise commitment.
What to Watch
What to watch
the V9-Medium mid-June beta target date, if it slips, that tells you something about the fine-tuning timeline. Also watch for an xAI DPA publication, which would remove the primary enterprise compliance friction.
The TJS read: the worktrees feature is a real workflow improvement for multi-thread agentic coding, don’t wait to test it if you’re on Grok Build already. On V9-Medium, xAI’s track record with stated timelines is worth consulting before adjusting roadmap plans around mid-June. And European teams: no DPA means no enterprise commitment until that changes.