Subscribe

AI · AltTab Dad

Grok Bot reset everyone's weekly limits after Elon promised overnight fixes

Quick answer: After paying users said their Grok Bot weekly usage was running out much faster than before, Elon Musk posted that the team was fixing token efficiency and, later the same day, that overnight engineering fixes were going in. On Nov 9, 2026, the official Grok Bot account, @bot, posted that it had reset weekly limits for all Grok Bot users. (Elon Musk, Nov 8; Elon Musk, Nov 8; @bot, Nov 9)

What’s new

A reset clears the weekly meter. It is not, by itself, a promise that the allowance is larger than it was last week. Check your own usage view (in Grok Bot under Settings → Usage & Billing, or at grok.com/dashboard on The X Subscription) for what your account actually shows. Official docs last updated Oct 11, 2026 still describe a shared weekly pool and do not publish new pool sizes for this reset. (Grok Bot FAQ; Usage limits)

What’s still unclear

Musk’s posts promised better token efficiency and smarter behavior, not a public changelog. This article does not independently test whether bots now use fewer tokens or finish tasks more reliably.

One widely discussed explanation for the quality drop, and for the later improvement some users reported, is that Grok Bot stopped routing as many tasks to Anthropic’s Claude Opus 5.5 and then started again. That specific on-and-off switch has not been announced by Musk, by @bot, by @SpaceXSI, or in the public docs reviewed for this piece.

The most detailed public version of the claim is a secondhand report, not an official changelog. On Nov 9, Joey Gilkey (@JoeyGilkey), CEO at TitanX, wrote that after meeting with the Grok Bot engineering team, “They admitted that it was almost entirely from pulling Opus 5.5 out. Which is now back as of this AM.” In the same post he said he was “somewhat surmising” about why quality collapsed over the weekend, including the idea that a mid-task model switch plus context compaction would have left in-progress work “incredibly dumb.” No engineer from that meeting has repeated the Opus admission in public. The closest staff comment on the record is from Sakthi Sankarraman (@ASankarraman), whose bio says he is building Cursor and @bot: on Nov 8 he replied to Gilkey’s earlier quality complaint, “Hi I work on @bot and I apologize for the quality not being up to par. Any chance we could chat what was not working for you, so we can make it better!” He did not mention Opus. (Joey Gilkey, Nov 9; Sakthi Sankarraman, Nov 8)

What is on the record: on Oct 7, 2026, Musk wrote that Grok Bot would use “the best back end model for any given task, including Claude Opus 5.5, MidJourney, Suno and other leading APIs.” The official FAQ still says users cannot pick a model, “cannot be routed to or away from a particular model,” and that “Grok Bot does not show the underlying model for a request.” It also notes that routing a task to Opus 5.5 does not change what you are charged per token; how fast you burn the weekly pool depends on the work and how much effort each task takes. (Elon Musk, Oct 7; Grok Bot FAQ)

Why it matters

Grok Bot is a paid desktop and phone assistant that can keep working in the cloud after you close the laptop. If the weekly meter empties in hours instead of days, routines stop unless you add extra paid usage. A company-wide reset lets people get back to work immediately. Whether the overnight efficiency fixes actually stretch that allowance, and which model is doing the work, still has to be judged from your own usage page and results. The product does not show the model name, so outside fingerprint tests remain unverified.

This is separate from Elon’s Nov 10 comments about Grok 4.8 training progress.

Sources

— AltTab Dad