SpaceX Closed the $60 Billion Cursor Deal. Grok Build Now Leads With Grok 4.7.

SpaceX closed its $60 billion purchase of Cursor on August 14, 2026, and by late October the Grok Build documentation no longer describes two coding engines: it leads with one model, Grok 4.7.

This guide first ran on August 10, when the deal was signed but not closed and Grok Build looked like a product with two engines under one name, so the open question was what would happen to that arrangement once the paperwork finished and the two teams could plan together. Eleven weeks later there is an answer. It is quieter than either camp predicted. Every claim below was re-read against a primary page on 27 October 2026, and one claim from the first version was wrong. We fix it below.

What changed since the August 10 version

Six things moved. Two of them overturn the original framing.

• The deal closed. Cursor posted "Cursor is now part of SpaceX" on August 14, 2026. Gigazine reported it two days later.
• The April terms were misstated. The first version called the $10 billion figure a payment for collaboration. TechCrunch describes it as a break-up fee if the buyout fell through.
• A new model arrived before the close. Grok 4.6 shipped on Wednesday, August 12. Cursor's post called it an early example of what the two companies can build together.
• Another followed. Grok 4.7 arrived on September 21, on a larger base than 4.6, according to Cursor's help page.
• Composer lives on in Cursor. The help pages for both Grok models say Composer is not being deprecated.
• The Grok Build docs went quiet on Composer. The overview page, read on 27 October, names no Composer option. It points at Grok 4.7 and at custom models set in a config file.

Re-checked 27 October 2026: the acquisition thesis holds, and the "two engines in one product" thesis does not. We cannot see inside every install, so read that last point as "the documentation no longer presents it". It is not proof that no user can still pick Composer.

The timeline in one table

The sequence matters more than any single headline. Dates are the ones the sources give.

The gap between signing and closing was 59 days. In that window the headline model changed twice.

How the deal came together

SpaceX's AI division is built around xAI and the Grok brand. Signs of interest in Cursor showed up early in 2026, when xAI hired two of the startup's senior engineering leaders. In April, a news report said xAI would rent out data center capacity to Cursor. Those talks turned into a deal announced the same month.

The April terms were an option, and a lopsided one. SpaceX could buy the company for $60 billion in stock, or it would pay a $10 billion break-up fee if the deal fell through. On June 16 it chose the buyout, days after its own IPO. TechCrunch said then that a close was likely in the third quarter. It came on August 14, inside that window.

The company SpaceX bought was Anysphere, founded in 2022. Cursor's own blog gives a sense of what it was building toward. In the May 18 post introducing Composer 2.5, it said it was training a much larger model from scratch with SpaceXAI, using 10 times more total compute. That is a statement about direction, written a month before the deal was signed.

What Composer 2.5 actually is

Cursor's changelog dates Composer 2.5 to May 18, 2026. It is built on the same open checkpoint as Composer 2, Moonshot AI's Kimi K2.5. Cursor says it trained the new version on 25 times more synthetic tasks than its predecessor. Standard pricing is $0.50 per million input tokens and $2.50 per million output tokens.

A bug report filed on June 9 against an unrelated open-source tool shows a paying subscriber saying Composer 2.5 had been released for Grok Build in early June, which is a week before SpaceX exercised its buyout option and well before anyone outside the two companies could say how the products would be combined. Odd timing. The model sat next to Grok 4.5, which SpaceXAI announced on July 8 as its most capable model so far.

One caution on figures. Earlier versions of this page quoted two benchmark scores for Composer 2.5. We could not find them on a page we can open, so they are cut. A score we cannot source does not belong here.

What Grok Build is, as of late October

Grok Build is a coding agent for the terminal. It is a separate product from the consumer chatbot. The overview page at docs.x.ai describes three ways to run it: an interactive full-screen terminal interface, headless mode for scripts and bots, and the Agent Client Protocol for use inside other apps.

Install is one command on macOS, Linux or Windows. On first launch it opens a browser to sign in, or it reads an API key from the environment. The page also documents a config file, ~/.grok/config.toml, where a developer can register any model that speaks a compatible API and make it the default.

Two details matter for this story. The page promotes Grok 4.7 as the company's latest model, available directly on the API. And it says nothing about Composer. A separate Cursor help page for Grok 4.6 calls that model first-party to Cursor and built together with SpaceXAI, the name those pages now use for xAI.

In short, the model that Cursor sells to its editor users and the model that the terminal agent leads with now share a lineage, built on the same shared training effort and the same data. In August they did not. That is the change.

Which model for which job

Cursor's own help pages draw the line between the models, and the line is about duration and cost.

Some specifics from those pages. Grok 4.7 has four effort levels (low, medium, high, xhigh), with high as the default. It offers a 256k standard context window and a 500k long-context option, with higher rates above 256k input tokens. On the Plus plan it is fixed at medium effort and Fast mode is off. Every paid plan includes it, and it works in the desktop app, the web, the CLI, the SDK, automations and the iOS app.

For a deeper look at how 4.7 scores against rivals, see our comparison of Grok 4.7, Opus 5, Astra and Gemini 3.8.

The money and the data angle

The headline number moved with the buyer's stock. The price was set in SpaceX shares. TechCrunch noted that the stock had climbed roughly 50% above its IPO price by pre-market trading on signing day, so the sum was a moving target. An all-stock price is a share count, not a fixed amount of cash.

Here is the money side in one place.

TechCrunch reports the round was on track to close with several large investors, Nvidia among them, before SpaceX made its move.

The data side matters as much. Per the Hokai record for Grok 4.7, xAI gave that model supplemental training on anonymized Cursor workflow data and taught it the runtime of the Grok Bot always-on agent. Fullstack makes a similar point about Grok 4.5, saying it incorporates data from real developer activity. An editor used by working engineers every day is a training signal that a chatbot cannot match. Arguably, that is worth more to the buyer than the revenue. Data wins.

Who it affects: the winners

Cursor subscribers get two lineages inside one editor: a fast, cheap model for routine edits and a heavier one for long jobs. They pay one bill. The Grok models draw from the same first-party usage pool as Composer, so switching between them does not open a second meter.

xAI gets a distribution channel it could not have built quickly, an editor with an installed base of professional developers. Cursor gets what its own post called access to the world's largest GPU fleet. That matters for a company that until recently built on someone else's open checkpoint.

Terminal users also gain. Grok Build can now be pointed at a model trained with coding data from a mature editor. For the always-on side of this family, read our comparison of Grok Bot, Meta Muse and OpenAI Dots.

Who it affects: the people who lose something

Nobody in the deal lost money. The people who lost something are developers who wanted a single, simple pitch.

The original Grok pitch was one lab, one model, one roadmap. Today there are four model names across two products, and two of them share a base. A buyer comparing Claude Code, GitHub Copilot, Windsurf and the Grok stack has to learn which name maps to which job before comparing anything else.

There is a second cost. Cursor said its policies would not change after the acquisition. Even so, a developer who chose Cursor in part because it was model-neutral now owns a tool whose parent builds a competing model family. That is not a defect today. It is a dependency that did not exist in July.

The other side of the argument

The strongest rebuttal to the "this is confusing" take is that nothing here is unusual. Offering several models inside one coding tool is standard practice. Several rival editors, among them Windsurf and Replit, already work this way. A router that offers a cheap model for small jobs and a strong one for big jobs is doing its job.

There is also a fair reading of the August arrangement as a trial run: Cursor's model appeared in a product it was about to join, before the paperwork closed. If that is right, the two-engine period was a transition, and the late-October docs show it ending with one headline model and a second one kept for cost. Neither company has said so.

Where this argument is weaker: the router comparison assumes a neutral party. A tool owned by a model lab has a reason to prefer its own models. In the two Cursor help pages we read, the advice is a first-party Grok model for hard work and Composer, another first-party model, for easy work. Neither page names a third-party option in that choice. That is a normal commercial position. It is not the same as a neutral menu.

How to choose today

Match the choice to the job and the editor, not to the brand.

1. You live in an editor. Stay in Cursor. Use Grok 4.7 for long tasks and Composer for quick edits.
2. Check effort first. The Plus plan fixes effort at medium, so a big run may need a higher plan.
3. You live in a terminal. Try Grok Build with Grok 4.7. Compare it with Claude Code on Claude Opus 5.5, using the same repository and task.
4. You want a neutral setup. Aider or the custom-model setting in Grok Build lets you bring your own endpoint.
5. You are still shopping. Read our ranking of AI coding assistants first. Then read the head-to-head of Cursor Composer 2 and Claude Code.

Whatever you pick, test on your own code, because vendor claims come from vendor suites and those suites rarely look like your repository, your language mix or your review habits. Vendor benchmarks use their own suites and conditions. Nobody outside the companies has published a table that scores Grok 4.7 and Composer 2.5 on the same suite on the same day. So the honest answer to "which is better" is a short trial with your repository.

For wider context, our guide to the best personal AI agents covers the always-on side, and Grok versus ChatGPT covers the chat products. The rival labs are Anthropic and OpenAI, whose GPT-5.5 is the other frontier option many teams compare.

What to watch next

Three things would change this picture.

• A same-suite benchmark. If xAI or an independent tracker scores Composer 2.5 and Grok 4.7 on one coding suite, the "which should I use" question gets a real answer.
• Composer's next release. Cursor says Composer is not being retired. A Composer 3 would show whether Cursor keeps a separate model line under SpaceX.
• Model-neutrality language. Watch whether Cursor keeps non-Grok models in its menu and how its help pages describe them. That is the cleanest test of whether the editor stays neutral.

We will re-read this page when any of the three moves. Until then, the summary is short. The acquisition is done, the headline model is Grok 4.7, and Composer survives as the cheap, fast option inside Cursor.

Frequently asked questions

What is Grok Build, and how does it differ from the Grok chatbot?

Grok Build is xAI's coding agent for the terminal. It runs as an interactive full-screen interface, headlessly in scripts, or through the Agent Client Protocol inside other apps. The Grok chatbot is a separate consumer product built for open-ended conversation.

Did SpaceX finish buying Cursor?

Yes. Cursor announced the completed acquisition on August 14, 2026 in a post titled "Cursor is now part of SpaceX". The deal was an all-stock transaction reported at $60 billion. Cursor said its policies would not change.

Should I pick Grok 4.7 or Composer 2.5 for coding?

Cursor's help pages point Grok 4.7 at hard, multi-hour tasks and Composer at everyday coding where speed and cost matter most. Nobody has published a benchmark that scores both on one suite on the same day. Run a short trial on your own repository before you commit.

Is Cursor retiring Composer?

Not according to Cursor. The help pages for both Grok 4.6 and Grok 4.7 state that Composer is not being deprecated and stays available for everyday coding speed and cost.

Can I use a different model inside Grok Build?

Yes. The Grok Build overview documents a config file, ~/.grok/config.toml, where you register a model by ID and base URL and set it as the default. You can then switch models inside the interface with the /model command.

Covered in this guide

• Cursor: Cursor is an AI code editor built on VS Code, widely adopted across large enterprise engineering teams, with Agent Mode, Tab completion, and Cloud Agents.
• Grok 4.7: xAI's September 2026 flagship model, built for coding and agentic work with a 500K token context window and a 46.3% CursorBench score.
• Aider: Free, open-source terminal pair programmer that edits your git repo and commits each change; bring your own LLM key. 49,000+ GitHub stars.
• Anthropic: Anthropic is a Public Benefit Corporation started by 7 ex-OpenAI researchers. It builds the Claude model family and counts Amazon and Google among its largest backers.
• Claude Code: Anthropic's agent for the terminal, IDEs, desktop, web and mobile that edits files, runs tests and opens pull requests. There is no free plan.
• Claude Opus 5.5: Anthropic's September 2026 flagship LLM with a 1 million token context window, built for long-running agentic coding and computer use.
• GitHub Copilot: GitHub's AI coding assistant with agent mode, Copilot Fusion model routing and MCP support across 6 IDEs, plus a free plan with no credit card required.
• GPT-5.5: GPT-5.5 is OpenAI's April 2026 flagship model for agentic coding and knowledge work, with a 1.05M-token API context window and text and image input.
• Grok: Chatbot from xAI that searches the web and X live, now sold with Grok Bot and Cursor inside The X Subscription.
• Grok 4.5: xAI's Grok 4.5 targets coding and agentic tool-calling workloads, built alongside Cursor and priced well under rival flagship models for high-volume agent loops.
• Grok 4.6: Grok 4.6 (Aug 2026) is xAI's 500K-context reasoning model, matching the top score on the Artificial Analysis Intelligence Index.
• Grok Bot: xAI's always-on AI agents that sign into your existing apps and finish multi-step jobs, included from the Premium tier up of xAI's combined subscription.
• Moonshot AI: Beijing AI lab founded in 2023 by Yang Zhilin, maker of Kimi and the open-weight Kimi K3 models, valued at $20B at its May 2026 round.
• OpenAI: OpenAI builds ChatGPT, the GPT-6 family (Astra, Sol, Terra, Luna), Codex and the OpenAI API. It reports 1.2B weekly ChatGPT users and raised new funding in March 2026.
• Replit: Cloud coding environment where you describe an app in plain English and AI builds and deploys it.
• Windsurf: Windsurf, now Devin Desktop from Cognition, is a VS Code-based editor with local and cloud agents, free with unlimited Tab and Pro at $20 a month.
• xAI: Elon Musk's AI lab behind the Grok models, now run inside SpaceX under the SpaceXSI name alongside Cursor, with a free-to-premium subscription that shares one usage pool.

Sources