Grok Bot will route tasks to Claude Opus 5.5, Musk says

Grok Bot will stop defaulting to xAI’s in-house Grok models and will hand each task to whichever back end suits it best, Elon Musk said on October 7. Anthropic’s Claude Opus 5.5 is first on his list. Musk wrote on X that SpaceX will use “the best back end model for any given task, including Claude Opus 5.5, MidJourney, Suno and other leading APIs,” and added: “Whatever is most likely to give you the best outcome.”
The same day, Lauren (@poteto), who works on Grok Bot at SpaceXAI, said it more directly: “all bots will be powered by opus 5.5, and they can spawn cloud agents of whatever cursor models we have.” Read literally, Opus 5.5 would be the main model behind every bot, and Cursor’s models would run as cloud agents beneath it. SpaceX now owns those Cursor models. It acquired xAI effective February 2 and bought Cursor in August in a $60 billion deal, and Grok Bot comes bundled with most SuperGrok and paid Cursor plans.
Grok Bot launched in beta on August 11. Each bot runs on its own cloud computer, keeps memory, works across connected apps and keeps going on tasks after the user closes the laptop. When SpaceXAI launched Grok 4.7 in September, it said the model was trained to “natively understand the Grok Bot harness”. Only weeks later, Lauren’s post named Opus 5.5 as the model behind all bots. Musk’s list names no exclusions. Testing Catalog, writing on Threads, still assumed OpenAI models would stay out of scope, and said the move raises questions about what to expect from upcoming Grok models.
Opus 5.5 came out on September 22. In Anthropic’s launch tests it scored 57.8 percent on CursorBench 4.0, against 51.8 percent for Claude Fable 5.1, and Anthropic says it costs 40 percent less to run than Opus 5. SpaceXAI has no music model, so Suno fills a gap. Midjourney would sit next to Grok’s own image generator, Imagine. Anthropic and SpaceX already do business with each other: on May 6, Anthropic announced a deal to use all the compute at SpaceX’s Colossus 1 data center, which it put at more than 300 megawatts and over 220,000 NVIDIA GPUs.
For people who pay for Grok Bot, Implicator.ai has the details that matter. Grok Bot has no model picker for customers. Usage analytics show which model served each request, including a replacement model used when the first choice fails, and billing follows the model that served the request. Musk hasn’t said when the routing starts or how a model is chosen for each task.
Replies to Musk’s post asked how privacy will be handled when Grok Bot conversations are sent to third-party APIs such as Anthropic’s, and one user asked for an opt-out, TeslaNorth reported. Musk did not elaborate.
Implicator.ai says enterprise teams get a model allowlist that Grok Bot honors by default, but enforcement isn’t guaranteed: during onboarding, customers see an acknowledgement that Grok Bot may not follow the list. Customers whose contracts limit subprocessors are told by existing security guidance to contact their account team before deployment.
Up next
Haiku 5.5 ranks #2 of 182 and burns 440M tokens doing it
415.30 seconds to first token: that’s Artificial Analysis’s reading for Claude Haiku 5.5 at max effort, versus a 2.17s tier median. Once it starts, though, it streams 243.4 tokens per second.
Related stories
- Grok Bot paid plans get 7% more usage
- Grok Bot exits beta and joins Slack, Teams and video calls
- Opus 5.5 matches Fable 5.1 on most work for less money
- Anthropic paid $740,000 to a firm led by Rubio’s ex-aide
- CNBC: Anthropic IPO likely to price above $70-78
- Microsoft shares slid as Claude got Sheets and Whiteboards
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

