Some free windows are worth catching. Right now, Grok 4.5 is available for free for a limited time in Grok Build, xAI’s terminal-based coding agent. This is not just a chatbot demo: Grok Build installs in your terminal, opens a TUI, reads a project, proposes plans, edits files and can run commands.
The public grok.com/build page routes you into the Grok Build flow. xAI’s CLI page says Grok Build is now powered by Grok 4.5, and the Grok 4.5 announcement explicitly says limited-time free usage is available in Grok Build and Cursor.
You need a Grok/xAI account, the CLI installer, then grok. The free access is temporary, so do not treat it as a permanent free plan.
What is free right now
xAI announced Grok 4.5 on July 16, 2026. The positioning is clear: coding, agentic tasks, software engineering and knowledge work. The official page says Grok 4.5 is available in Grok Build, Cursor and the xAI console, with public API pricing at $2 / million input tokens and $6 / million output tokens.
The interesting part is the free window: xAI says Grok 4.5 usage is free for a limited time in Grok Build and Cursor. The TUI screenshot also shows the message: “Grok 4.5 is here, try it out for free for a limited time!”
Install Grok Build
Open grok.com/build, create or sign into your account, then copy the command for your operating system.
On Windows, the PowerShell tab shows:
irm https://x.ai/cli/install.ps1 | iexOn Linux, macOS or WSL, xAI’s docs show:
curl -fsSL https://x.ai/cli/install.sh | bashAfter installation, open a terminal in a test folder and run:
grok
Start a session
On first launch, Grok Build asks you to authenticate. Once signed in, the interface offers a new worktree, session resume, changelog, or a normal interactive session. As with Claude Code, Codex, OpenCode or ZCode, start on a copy of a project or a dedicated branch.
Grok Build is designed for complete workflows: plan mode, subagents, skills, hooks, MCP servers, memory, code search, terminal execution, headless tasks and code review. The point is not only to ask questions, but to let the agent work inside the repository context.
Why it is worth testing
Grok 4.5 comes with an aggressive engineering pitch. xAI highlights strong software-engineering results, serving around 80 tokens per second, and better token efficiency than some competing models. It is also the default model in Grok Build.
In the current coding-agent landscape, it is worth a real field test. We have already seen useful free windows with ZCode and GLM-5.2, and Qwen 3.6 Plus in OpenCode. Grok Build adds another terminal-first option next to Claude Code and Codex.
The privacy caveat
I would not publish this without a strong privacy note. Grok Build recently faced criticism after analyses suggested the tool uploaded full Git repositories to xAI infrastructure. The Verge reported that xAI disabled the feature and that Elon Musk promised deletion of previously collected data. xAI also promoted the open sourcing of Grok Build.
That does not mean you should avoid testing it. It means you should test it cleanly: not on a confidential client repository, not in a folder with secrets, and not on a repo with API keys committed in history. Start with a clean worktree, demo repo or disposable folder.
My take
Grok 4.5 being free in Grok Build, even temporarily, is a strong opportunity for a hands-on benchmark. The install path is simple, the TUI is already capable, and the model is clearly aimed at complex coding work.
My recommendation: try it now, but keep strict hygiene. Create the account, install the CLI, run it on a test project, inspect the plans, diffs, speed and command behavior. If the result is good, you have another card in your coding-agent stack. If the free window closes, you will at least have evaluated Grok 4.5 without paying.
