Musk: Grok Bot will route tasks to Claude Opus 5.5, Midjourney and Suno

In a notable reversal of xAI's in-house-only posture, Musk confirmed Grok Bot will send tasks to whichever model is best for the job. That means Anthropic's Claude Opus 5.5 for complex reasoning, Midjourney for images and Suno for audio. Grok Bot has been in beta since mid-August and reports 418,000 weekly users. It previously ran only on xAI models like Grok 4.5, priced at $2/$6 per million input/output tokens.

The agent becomes an orchestrator. A routing layer classifies each request and dispatches it to the best-fit model, with compute drawn from SpaceX's 300MW Colossus datacenter, according to aggregator reports. Alongside the change, Grok Bot v0.68.1 added native presentation-slide generation.

The move fits the week's theme of routing going mainstream. Bedrock and Microsoft Foundry already host rival labs' models side by side. Microsoft's Copilot now hands work to local models, and AWS's Strands Decider 2B is built to make routing decisions. For xAI (now reportedly SpaceXAI), admitting a competitor's model is better at reasoning is a pragmatic but striking concession.

Open questions: the commercial terms with Anthropic, whether users will see which model answered, and data-handling implications when queries leave xAI's stack. Security researchers also say xAI ignored bug-bounty follow-ups for months. Routing user data across more vendors expands the trust surface.