Grok Build is xAI's entry into agentic coding, and it arrived swinging. Launched May 14, 2026, it's a terminal-native agent that runs up to eight subagents in parallel, each isolated in its own Git worktree, backed by the new grok-build-0.1 model. It's fast, it's cheap if you already pay for SuperGrok, and the worktree architecture is genuinely clever. But after a week of real work, the honest read is that the model itself still trails Claude Code and Codex on quality.
Rating: 4.0/5
Verdict: Grok Build is a strong first release with a smart parallel-worktree design and excellent value for xAI subscribers. The grok-build-0.1 model is the ceiling — good, not category-leading — so it's a fast second agent rather than a Claude Code replacement.
Best for: xAI subscribers who want a cheap, fast terminal agent for parallelizable work and don't need the absolute best code quality.
What is Grok Build?
Grok Build is xAI's agentic command-line coding tool — the company's direct answer to Claude Code and OpenAI's Codex. You give it a task in the terminal, it drafts a plan for your approval, then edits files, runs commands and returns diffs you can review. It's terminal-first by design, aimed at developers who live on the command line.
The headline architecture is parallelism: Grok Build can dispatch up to 8 subagents at once, and each one works inside its own isolated Git worktree. That means one agent can refactor a module while another writes tests and a third updates docs, with no file conflicts between them. It's MCP-compatible too, so external tools and data sources plug in cleanly. It sits alongside the leaders in our best AI coding assistants roundup.
The model and benchmarks
Grok Build runs on grok-build-0.1, xAI's first coding-tuned model, with a 256K-token context window. On benchmarks it scores 70.8% on SWE-bench Verified — a respectable number for a "0.1" release, but roughly 15–18 points behind Claude Opus 4.8 and GPT-5.5, which sit in the high 80s.
That gap shows up in practice. On self-contained tasks — add an endpoint, write a component, fix a well-scoped bug — Grok Build is quick and usually right. On the gnarly stuff, where a change ripples across many interdependent files, it makes more mistakes and needs more supervision than the leaders. For cost-sensitive agentic work xAI also offers Grok Code Fast 1, a cheaper, faster model tuned for high-volume loops.
Parallel subagents and worktrees
The isolated-worktree design is the best idea in the product. Each subagent gets a clean checkout, so eight agents can run simultaneously without clobbering each other's files. For a batch of independent tasks, that compresses an afternoon of sequential work into minutes — fire off the plan, watch the worktrees fill in, review the diffs as they land.
The catch is coordination. Because each agent is isolated, tightly coupled changes — where subtask B genuinely depends on subtask A's output — don't benefit and can actually get messier, since the agents can't see each other's in-progress work. Grok Build's sweet spot is the parallel grind, not the delicate cross-cutting refactor. We dig into that tradeoff in Grok Build vs Claude Code.
Grok Build pricing in 2026
Grok Build isn't a standalone subscription — access rides your xAI plan, with higher limits at each tier. There's also a straight API rate if you'd rather pay per token.
| Plan | Price | What you get |
|---|---|---|
| SuperGrok | $30/mo | Standard Grok Build access and limits. |
| X Premium+ | $40/mo | Grok Build plus X (Twitter) Premium+ perks. |
| SuperGrok Heavy | $300/mo | The highest limits for power users. |
| API — grok-build-0.1 | $0.20 / $1.50 | Per million input / output tokens. |
| API — Grok Code Fast 1 | $0.20 / $0.50 | Cheaper, faster model for agentic loops. |
For anyone already paying for SuperGrok at $30/mo, Grok Build is effectively free upside — that's the real value story here. The API rates are aggressive too: at $0.20 in / $1.50 out per million tokens, grok-build-0.1 is meaningfully cheaper than running Claude Opus through the API. Full breakdown in our Grok Build pricing guide.
Pros and cons
After a week of real tasks across the CLI, here's the honest ledger.
Pros
- Smart 8-way parallelism with isolated Git worktrees
- Excellent value bundled into $30/mo SuperGrok
- Cheap API rates ($0.20 in / $1.50 out)
- Fast, plan-first loop with MCP support
- Strong on self-contained, parallelizable tasks
Cons
- grok-build-0.1 trails the leaders by 15–18 pts on SWE-bench
- Weaker on interdependent, cross-cutting refactors
- 256K context is smaller than some rivals
- Ecosystem is young — fewer integrations and docs
- Best limits are gated behind $300/mo Heavy
Who Grok Build is for
Pick Grok Build if you already pay for SuperGrok and want a fast, cheap terminal agent for parallelizable work — batches of independent tickets, scaffolding, test-writing and the like. The worktree model genuinely shines there, and the price is unbeatable when it's bundled with a plan you already have.
Look elsewhere if code quality on hard problems is your top priority. Claude Code and Codex both out-model grok-build-0.1 by a clear margin, and OpenCode gives you full model choice. In practice, the smart move for xAI subscribers is to run Grok Build as a fast second agent and reach for Claude Code when the stakes are high.
As a first release, Grok Build is impressive and earns a solid 4.0/5. The architecture is ahead of the model — once grok-build-1.0 closes the quality gap, this becomes a much harder tool to ignore.
Frequently Asked Questions
What is Grok Build?
Grok Build is xAI's terminal-native agentic coding tool, launched May 14, 2026. It runs up to 8 parallel subagents, each in its own isolated Git worktree, and is backed by the grok-build-0.1 model with a 256K-token context window. It competes with Claude Code, Codex and OpenCode.
How good is grok-build-0.1?
grok-build-0.1 scores 70.8% on SWE-bench Verified — respectable for a "0.1" release but roughly 15–18 points behind Claude Opus 4.8 and GPT-5.5. It's strong on fast, parallelizable work and weaker on the very hardest interdependent refactors.
How much does Grok Build cost?
Access is bundled into xAI subscriptions: SuperGrok at $30/mo, X Premium+ at $40/mo, and SuperGrok Heavy at $300/mo for the highest limits. Via the API, grok-build-0.1 runs about $0.20 per million input tokens and $1.50 per million output, with the cheaper Grok Code Fast 1 at $0.20/$0.50.
What makes Grok Build different?
Its signature feature is 8 parallel subagents, each running in its own isolated Git worktree, so agents can work on different parts of a codebase simultaneously without stepping on each other. It also ships a plan-first execution loop and MCP support.
Is Grok Build better than Claude Code?
Not on raw code quality — Claude Code's Opus 4.8 leads benchmarks by a wide margin. Grok Build wins on parallelism and price for xAI subscribers. Most quality-first developers still pick Claude Code; see our Grok Build vs Claude Code comparison.
Does Grok Build support MCP?
Yes. Grok Build is MCP-compatible, so you can connect external tools and data sources, and it uses a plan-first loop where you approve the plan before it executes. That puts it in the same class of agentic surface as Claude Code and Codex.