METAGROK
GROK BOT / which-model-does-my-grok-bot-use

Which model does my Grok Bot use?

which-model-does-my-grok-bot-use

Finish this page knowing that Grok Bot picks a model for each task, that you cannot choose or block one, and where two official pages disagree on whether the name is shown and how that choice is billed.

This is a Grok Bot question, not a Grok chat-app question. On legacy Cursor plans, usage is metered on your Cursor account. On The X Subscription it draws from the shared weekly pool shown under Usage at grok.com/dashboard. More Grok Bot jobs live on the Grok Bot hub under Usage, plans and billing.

What the official pages say

Cursor help (Models) is the product spec for how Grok Bot routes work:

  • Grok Bot is not tied to a single model. For each task, it uses the backend model most likely to deliver the best outcome. That can be a first-party xAI model or a third-party model from a provider on the sub-processor list.
  • You cannot see which model handled a request. Usage and spending views record the activity as Grok Bot usage, not by provider or model name.
  • You cannot choose, request, or block a specific model. There is no per-user or per-request model selection, and there is no model picker.
  • Routing a task to Opus 5.5 does not change what it costs you per token.

xAI's security page (Models and data) agrees that Cursor manages model selection, that there is no customer-facing model picker, and that the serving mix can change over time with no fixed vendor set guaranteed. On two other points it disagrees with Cursor help. Name both sides; this page does not pick one.

  1. Can you see the model? Cursor help says Grok Bot does not show the underlying model for a request. The xAI security page says usage analytics show the model that served each request, including failovers.
  2. Does the model change the cost? Cursor help says routing to Opus 5.5 does not change what it costs you per token. The xAI security page says billing follows the serving model.

A third official disagreement sits next to “can I block a model?” Cursor help says Grok Bot is not governed by your team's Cursor model allowlist. The xAI security FAQ says the allowlist is Enterprise only, and enforcement is not guaranteed; the list is honored by default.

Nothing in those pages describes linking your own Anthropic, Claude, Midjourney, or Suno subscription so Grok Bot will use it. Cursor help meters Grok Bot usage on the Cursor account.

What an executive said, separately

On Oct 7, 2026, Elon Musk wrote that Grok Bot “will use the best back end model for any given task, including Claude Opus 5.5, MidJourney, Suno and other leading APIs” (the post). That is commentary from an executive, not a product spec. Cursor help and the xAI security page do not name Midjourney or Suno as live Grok Bot tools. Asking in chat for a named model is allowed; nothing documents that the ask will be honored.

What you can actually do

  1. Open Grok Bot and give the Bot the task. Do not look for a model menu. There is not one.
  2. Read spend as Grok Bot usage. On legacy Cursor plans, that is Settings > Usage & Billing. On The X Subscription it draws from the shared weekly pool shown under Usage at grok.com/dashboard. See Cut Grok Bot weekly usage when the meter climbs fast.
  3. If two official pages disagree on visibility or billing, keep both in view and check the live pages again later. Do not treat a forum reply or an @grok answer as the tie-breaker.

For how the weekly meter and on-demand cap work after the model has already been chosen, see Understand Grok Bot usage and “Bot failed to respond” and Control Grok Bot On-Demand after weekly usage hits 100%.

Pitfalls

  • Looking for a picker, a per-request override, or a way to force one vendor. Official help says those do not exist.
  • Reading the Oct 7 executive post as a list of models you can turn on today. The help pages do not name Midjourney or Suno.
  • Assuming Grok Bot draws on your own Claude plan. On legacy Cursor plans, usage is metered on your Cursor account. On The X Subscription it draws from the shared weekly pool shown under Usage at grok.com/dashboard.
  • Treating one official page as the only word on model names and cost. Cursor help and docs.x.ai still disagree on those two points as of Oct 19, 2026.