Image: SpaceXAI

Update: We now know more about how this works. Lauren (@poteto), who works on Grok Bot at SpaceXAI, says all bots will be powered by Claude Opus 5.5. They'll also be able to spawn cloud agents running any of Cursor's models. That answers a question many readers had: Opus 5.5 isn't just an occasional pick for certain tasks, it's the main brain behind every bot.

all bots will be powered by opus 5.5, and they can spawn cloud agents of whatever cursor models we have

— lauren (@poteto) October 7, 2026

Grok Bot is about to stop being a Grok-only product. Elon Musk says SpaceX will now run each task on whichever AI model does it best, even when that model comes from a rival.

Important note regarding Grok @Bot:

Going forward, @SpaceX will use the best back end model for any given task, including Claude Opus 5.5, MidJourney, Suno and other leading APIs.

Whatever is most likely to give you the best outcome.

— Elon Musk (@elonmusk) October 7, 2026

The examples he picked say a lot. Claude Opus 5.5 is Anthropic's top model, which recently landed in Google's Antigravity too. Midjourney is the best-known name in AI images, and Suno is the leader in AI-generated music. None of them is made by SpaceXAI.

Grok Bot is SpaceXAI's always-on agent. Each Bot gets a computer of its own, connects to your apps and keeps working on tasks after you close your laptop. Until now, the natural assumption was that it ran on SpaceXAI's own Grok models.

That assumption had good reason behind it. When SpaceXAI launched Grok 4.7, it said the model was trained to "natively understand the Grok Bot harness." Musk's post doesn't say Grok 4.7 is going away, only that it will no longer be the default answer to every job.

SpaceXAI's own numbers help explain the change. In its Grok 4.7 launch table, Grok 4.7 beats every rival on legal work and electrical engineering. But on multi-hour terminal work, it scores 37.6% against 57.9% for Anthropic's Fable 5.1. No single model wins everything, and SpaceXAI's chart shows it.

Image and music are even clearer gaps. Grok has its own image tool, Imagine, but Musk is effectively admitting that for some jobs, users will get a better result from Midjourney. And SpaceXAI doesn't have a music model in its lineup at all, so Suno fills a hole rather than replacing anything.

The Anthropic angle is the most interesting. The two companies are already partners: SpaceXAI rents Anthropic its Colossus 1 supercomputer, more than 220,000 Nvidia GPUs, to boost capacity for Claude Pro and Max subscribers. Sending Grok Bot tasks to Claude makes SpaceXAI a customer as well as a supplier.

For users, this is mostly good news. You hand a task to your Bot and it picks the tool, the way a manager hands work to whoever on the team is best at it. It also makes Grok Bot look less like a model showcase and more like a product that competes on results.

It's an unusual pitch for an AI lab. Grok Bot is now openly mixing in competitors' models, a bet that being the best agent matters more than having the best model.

Plenty is still unclear. Musk didn't say when the routing starts, whether users can see or choose which model handles a task, or whether outside models will cost extra or eat into plan limits faster. Opus 5.5 is one of the priciest models around, so that last question matters.

There's also the privacy side. Sending a task to Claude, Midjourney or Suno means your data goes to those companies, and enterprise customers will want to know exactly what's shared and under which terms. SpaceXAI hasn't published anything on that yet.

Speaking for myself, I love this. Picking the right tool for each job is exactly what an agent should do, and I think it's going to be a game changer.

For now, it's a single post from Musk rather than a product page. Once SpaceXAI posts the details, the real test will be whether Grok Bot tells you which model did the work.