Musk Reset Every Grok Bot Weekly Limit — Here Is What Actually Changed

Grok Bot subscribers spent the first week of November watching weekly allowances empty in days. xAI has now reset limits for everyone and Musk has said fixes are on the way. Here is what is confirmed, what each plan actually includes, and the part of the billing model that explains why the meter moves faster than people expect.

What each plan costs and includes

Read off xAI’s own pricing documentation on 10 November 2026. Multipliers are relative to the Plus base allowance, so Super is roughly 3.75× Premium and Ultra roughly 12.5×. Prices exclude sales tax. Grok Bot is not available on Free or Plus.

The multiplier is the thing to read, not the price. Ultra costs 6.7× Premium and gives 12.5× the allowance, so per unit of usage the expensive plan is the cheaper one. That matters because the complaints came disproportionately from Ultra subscribers — the people with the most allowance were the ones running out.

How the week unfolded

Why your allowance drains faster than it looks like it should

This is the part worth understanding, and it is in xAI’s documentation rather than in any announcement.

• Usage is counted in agent steps and tokens, not messages. Ten short questions may cost less than one long task. There is no per-message quota to budget against.
• The pool is shared. On The X Subscription, Grok Bot draws from the same weekly pool as the Grok apps and Cursor. Heavy use in one place reduces what is left for the others.
• You cannot choose the model. Grok Bot selects one per task and the choice can change between requests. It may be a first-party model or a third-party provider, and the model used is not shown to you.
• Model choice does not change the per-token price — but the amount of work a task requires does, so two requests that look similar can cost very differently.
• Linked accounts do not stack. If a SuperGrok subscription is linked, Grok Bot uses whichever source grants more weekly usage, not the sum. An upgrade can take up to 24 hours to register.

Put together, that is a billing model where the thing you are charged for is invisible at the moment you trigger it. Which is a reasonable design for an agent and a difficult one to budget against — and it explains the gap between what subscribers expected and what the meter showed.

How to keep control of it

• Check usage before a big job, not after. The Usage section of the Grok dashboard shows what is left in the weekly pool.
• Scope tasks tightly. xAI’s own guidance is to start small and keep jobs narrow. A single large task can consume a surprising share of an allowance in one run.
• Decide about overage deliberately. On The X Subscription, prepaid credit covers usage beyond the pool. Turn off Auto Top-up, or set a monthly additional usage limit, if you do not want that to happen automatically. On legacy plans the equivalent is on-demand billing through Cursor.
• Know that a spending cap is not a hard stop. xAI states that a cap does not halt a task already running, so a job in progress can push slightly past it.
• If you are consistently running out on Premium, the multipliers mean upgrading buys more allowance per dollar than it costs. That is the opposite of how most subscription ladders work.

What has not been answered

Worth being clear about the gaps, because the announcement left several.

• No timeline for the token-efficiency fixes beyond “underway”.
• No statement of what the fixes change, or by how much.
• No confirmation that weekly allowances themselves are going up — a reset refills the existing allowance, it does not enlarge it.
• No figures behind the cost claim, and no confirmation of how far limits were tightened in the first place, or whether they were tightened at all rather than demand simply rising.
• No change to pricing announced.

So the reset is real and immediate. The fix is a stated intention. Those are different things, and anyone deciding whether to upgrade this week should treat them differently.

Frequently asked questions