New Grok Bot user here and seeking advice on how best to manage Cursor tokens and usage expense after requesting the account usage explanation shown below. Have never used Cursor before for any purpose.
BIG WOW – used 1.25 BILLION tokens during the first week while just getting familiar with Grok Bot and spinning up a handful of new bots doing relatively simple tasks. Apparently many here on X sharing cool Grok Bot use cases are spending A TON of money to do so.
Found out the hard way. Jumped right in last week after studying the early success of others and spun up four or five bots doing relatively straightforward tasks. Within two or three days had reached the usage limit of the $60 entry-level Pro+ subscription. Added a $200 Ultra subscription and the (supposed) 20X capacity was used up within the next two or three days.
Worse, Grok Bot shuts off the ability to use the system after hitting an arbitrary weekly usage limit after day 5. Functionality and potential use cases are off the charts. Value proposition for new users sucks big time. Now I'm dead in the water and can do nothing more for a while.
Difficult to understand how to use it in our business if it gets turned off arbitrarily every two or three days...
================
I'll look up how Cursor's included usage and billing screens work so I can explain this in plain language. This screen is not an extra bill. It is a breakdown of AI usage that is already covered by your Cursor subscription for the current billing month (Aug 25–Sep 25, 2026).
A token is a small chunk of text the AI reads or writes (parts of words, code, punctuation). Big chats, big files, and long agent runs use more tokens. “M” means million.
The Usage column is 0.0% on every row. In plain terms: this work is still inside your included monthly allowance. You are not being charged extra for it on this screen.
Cursor splits included usage into two buckets.
________________________________________
1. Cursor Models — 524.4 million tokens
These are Cursor’s own models (the Grok family it sells as first-party).
What you usedAbout
cursor-grok-4.6-high-fast164.1M — newest Grok, “high” quality, fast (usually costs more per token than the non-fast version)
cursor-grok-4.5-high-fast137.1M — previous Grok, high + fast
cursor-grok-4.6-high118.2M — newest Grok, high quality, normal speed
cursor-grok-4.5-high105M — previous Grok, high, normal speed
This is mostly you (or agents) coding in the editor with Grok selected.
________________________________________
2. Other Models — 720.8 million tokens
This bucket is for everything else: third-party models, plus some Cursor features that get listed here even when they have their own allowance.
Biggest items:
•sand-automation (431.5M) — Grok Bot routines (scheduled / repeating jobs). These often look huge because each run can resend a long conversation and count cached context, not just new words. Cursor staff say this comes from Grok Bot’s own pool and should not eat your Claude/GPT coding quota.
•claude-opus-5-low (119.4M) — Anthropic Claude Opus, “low” thinking/effort setting.
•sand-default (86.5M) — normal Grok Bot chats. Same idea: listed here, metered separately from your main coding pool.
•cursor-grok-4.5-high-fast / cursor-grok-4.6-high-fast under Other Models — more Grok usage that landed in this list (overflow after the Cursor Models pool, or how the dashboard groups rows).
•sand-cua (16.9M) — another sandbox/agent-style feature (computer-use / cloud-agent style work).
•claude-opus-5-thinking-low (1.6M) — Opus with a thinking mode, low setting.
•gemini-2.5-flash (3.5M) — Google’s cheaper/faster Gemini.
________________________________________
What this means for you
•You used a lot of tokens this cycle (about 1.25 billion added together).
•Most of that is Grok in the editor plus Grok Bot automation, not a pile of expensive Claude/Gemini coding.
•0.0% everywhere means this page is still “included.” Extra charges, if any, would show under On-Demand Usage, not here.
•Unused included usage does not roll over when the cycle resets on Sep 25.
If the huge sand-automation number surprises you, check Grok Bot routines and pause or delete ones you don’t need. That is usually what inflates that line.
This.
I found out the hard way during the past week. Jumped right in and spun up four or five bots doing relatively straightforward tasks and within two or three days I had reached the usage limit of the entry-level pro plus subscription.
Added an ultra subscription and that capacity was used up within the next two or three days. Functionality and potential use cases are off the charts. That proposition sucks big time. How can I use it in my business if it gets turned off arbitrarily every two or three days.
@kgu87@JasonL_Capital I guess time will tell. I named my new team member Trader. He placed his first trade this morning and so far he’s up a few bucks….
@Teslaconomics Have already tried it with good outcomes so far. Allowed the bot to use a dedicated bank account to send funds with Link. Allowed it to charge credit cards and deposit funds using Stripe. Big wow.
Great photo! Doing well but missing aviation after owning Bonanzas and Malibus for about 20 years….
Unfortunately, after moving to the beautiful great Smoky Mountains a few years ago, we learned that our regional general aviation airports have no hangers available and have 10 to 20 year waiting lists.
Making matters worse, none of them have any plans to build hangers and will not approve private construction projects for new hangers. Might be grounded for a while - no need to get a new Sky ride if we can’t keep it inside…
A back office person working at one of our brokerage firms has over 30 years of experience with different IPOs and different sponsoring brokerages.
During our prep to get allocations from two different firms this past week she shared the harsh reality of how these things usually work, regardless of whether or not SpaceX may have allocated X percent for retail investors.
At the end of the day, according to this person, each brokerage decides who gets an allocation and the actual allocation amount. It is done the good old-fashioned way, based on relationships and the sole judgment of the brokerage.
Big All-In pod fan here, always enjoy the show!
We have already started this fun journey with our CyberLift venture. Now serving the beautiful mountain regions of NC, GA, TN. Using a growing Model Y startup fleet to start the music with safety drivers until Cybercab and RV are available for fleet orders and autonomous regulatory approval arrives in our markets.
We setup an interim enterprise IT infrastructure with mobile apps, cloud servers, payment gateways, etc. to hold us over until the Robo app network goes live across the US sometime down the road.
Our system works anywhere in the US. Happy to consider options for teaming up on the effort!
@truthandtesla That's fine - whatever works best for you. Wishing you great success. We have hundreds of happy CyberLift customers, a growing fleet, and more users coming online every day. Might be able to share some tips and save you some time/$ if you ever feel like a chat.
That's fine - whatever works best for you. Wishing you great success. We're already up to hundreds of happy CyberLift customers and growing every day. Might be able to share a few tips and save you some time and $ if you ever want to chat.
@CollinJHumphrey Today's factory-built homes are beautiful! As an example we offer may of these for under $300K (including land with public water/sewer/power/broadband) in a beautiful mountain community: https://t.co/YGzTEZRHla
FWIW - just sent this question to our fleet rep at Tesla.
By now I’m sure you’ve seen the growing number of online questions from Tesla owners regarding whether or not a vehicle equipped with the $8K lifetime FSD license will be able to operate in an unsupervised manner with no additional licensing or subscription cost when the FSD software evolves to that point and the required regulatory approvals are in place.
A related question is whether or not such FSD-equipped vehicles acquired and active before 2/14/26 will be able to participate in the Tesla Robotaxi network in the future with no additional licensing or subscription cost as previously announced by Tesla executives.
These questions are directly applicable to CyberLift operations as we are preparing to invest another $______ to acquire more vehicles with lifetime FSD, add the lifetime FSD license to our existing fleet vehicles, and join the Robotaxi network when this option becomes available in the future.
We are planning these moves because until now there has never been any question in our mind that once an owner acquires a lifetime FSD license, it stays with the vehicle for its lifetime and entitles the owner to receive all future FSD updates and capabilities that can be supported by the vehicle hardware, including unsupervised FSD and Robotaxi participation.
Can you please clarify these questions for us?
As always, thanks for your great support as we continue to grow the CyberLift business!
Very possible and definitely a concern. This potential scenario is about the only remaining reason we have not yet acquired the permanent FSD license for our fleet vehicles. All part of the calculated risk equation related to the purchase decision at hand.
At the moment, the odds of substantial subscription price increases seem much higher than the odds of totaling a vehicle before the end of its useful life, but one never knows for sure….