I was actively making calls with Dots, but I ended up hitting the limit. Given that I haven’t seen similar reports from others, it’s possible that making calls consumes the GPT-6 Astra allocation at the same rate.
Since executives such as Tibo have said that the limits are generous, calls may consume a very large amount of the quota.
You’ve hit the preview limit. Try again at 4:10 AM.
It seems the rate resets daily. Since I was able to talk for over four hours, I have no complaints about the generous terms
Same here, but actually without calls.
My first message to my Dot was at 30/9 at 9:47 AM Israel time. That was 2 days ago.
For the past 2 days, at 1:48 AM my dot would stop replying, showing a red forehead animation in the avatar, only starting to reply again at 9:48 AM.
Now I understand why. Its a shame that this limit is not documented and theres no way to see usage.
Adding another data point that is not the long-call case.
I hit a Dot-specific stop during the first week after launch. The UI said:
“Your Dot is resting. Congratulations on being one of the most active users! You have reached our usage limit to prevent abuse. Come back in a while!”
The message did not say that my normal plan allowance had run out. This appears to be an additional usage or anti-abuse limit affecting Dots themselves, and I haven’t been able to find any public documentation explaining how large that limit is or how it is measured.
The launch wording makes this especially confusing. OpenAI says Dot usage won’t count toward eligible users’ normal plan allowances during the first month. I am not saying OpenAI promised mathematically unlimited usage. The issue is that the public explanation doesn’t make this additional stopping point clear.
I hit it within the first week, and there is no visible meter showing how close I am to it or, in my case, a clear explanation of what workload caused it.
It would be very helpful if OpenAI documented:
Right now, “doesn’t count toward your plan allowance” and “has no additional usage ceiling” are clearly two different things, but the product doesn’t make that distinction very visible.
Hi @GGA981 , How long did your dot stay on cooldown? When does it reset?
This time the main chat was stopped for under two and a half hours, and the message gave no reset time.
On Oct 5, Taipei time, this was already on screen at 13:05:
“Your Dot is resting. Congratulations on being one of the most active users! You have reached our usage limit to prevent abuse. Come back in a while!”
The same message was still there at 14:30. A normal reply was possible at 15:01. I can’t say when it resets, or that it resets at the same clock time every day.
The task itself had already been open for about 116 hours by then. That is elapsed time, pauses included. It is not a measurement of how large this cap is.
It seems that dots can spawn around six multi-agents at a time, and I have a hypothesis that giving dots a large number of tasks causes it to spawn many multi-agents, which may consume the token allocation
This is also just speculation on my part, but using Computer Use 24/7 may be about as token-inefficient as voice calls
An OpenAI executive has stated the following:
Because usage on your primary dot is virtually unlimited at the moment, I can’t really give a reset. I need to come up with something new fast. (quoted verbatim)
I think the phrase “virtually unlimited at the moment” makes it sound as though usage is effectively unlimited, which is misleading and extremely confusing.
In reality, there are abuse limits enforced over rolling time windows, as well as a preview limit that resets every 24 hours. They advertise it as if there are no ropes here, but in reality, it feels like being tied up with ropes.
This is still only speculation, but it seems that this preview limit varies depending on the tasks being allocated or the user account. Some people apparently cannot experience 24-hour operation even without using calls
I don’t know what is causing this
@rabbitdaisuke Second stop, same message, still no clock.
The main chat could reply again at 15:01 Taipei on Oct 5. At 11:07 on Oct 6 it was resting again. That is about 20 hours after it recovered, not 24 hours after the first stop or the recovery. The message still gave no reset time.
Just to confirm, aren’t you subscribed through Google Play or the Apple Store?
In my experience, OpenAI clearly discriminates against users who subscribe through these platforms in favor of users who subscribe by credit card
I subscribe by credit card, and I am able to run dots for 24 hours in a multi-agent configuration
@rabbitdaisuke Credit card, not Google Play or the App Store.
This account still hit the same resting message about 20 hours after the previous recovery, and the message still gave no reset time. A store-subscription restriction does not explain this stop.
@rabbitdaisuke The second stop ended. It was longer than the first, and the only clock that fits is UTC.
The resting message was still on the main chat at 15:00 Taipei on Oct 6. A normal reply was possible at 15:05. From 11:07, when that message was already showing, that is 3 hours 58 minutes. The first stop was under two and a half hours.
A tool notice said “7:01 AM” and did not name a timezone. 7:01 AM UTC is 15:01 in Taipei, which falls between those two probes. Read as Taipei time, 7:01 AM does not match this recovery.
Just to confirm, how much are you using Computer Use?
I need to determine whether your dot has a lower usage limit allocated to it, or whether you are simply using Computer Use too much
By the way, I have now ported MCP entirely to headless software, so with moderate Computer Use and sub-agents (which may be consuming Codex usage), I have set a new record for my longest continuous runtime three days in a row
Hi @rabbitdaisuke, - hitting a limit in the middle of your work is frustrating.
According to the Dots documentation, conversations with your dot don’t count toward your ChatGPT usage limits. Tasks it starts or manages in Work or Codex count toward those products’ usual limits. Your plan also includes an allowance for deeper work, with extended limits for the first month after launch.
Dots can also research in the background, review connected information, and run scheduled tasks, as explained in the Dots guide. These guides don’t specify exact limits or how every activity is counted, so they don’t establish what triggered the limit you’re seeing.
For account-specific clarification, or if you still see the limit after the displayed retry time, please contact Support through the chat bubble at help.openai.com. Include the exact message and relevant timestamps, including your timezone.
~ Saisha
Your response was written without understanding the context of this discussion. I’m sorry to say, unhelpful
I was interested in whether staying on calls 24/7 actually causes users to hit account-specific limits or preview limits, and I was trying to discuss the possibility that these limits vary from account to account. I was in the middle of asking other participants about this
Replies from OpenAI Support are always welcome, but you suddenly came into this discussion and told me, “Our dictionary of laws doesn’t mention these limits. Therefore, we cannot help you.” And, “According to our laws, conversations with your dot don’t count toward your ChatGPT usage limits”
It feels like a thermal power generator ran out of fuel and shut down, only to be told, “The law says that thermal power generators need fuel, but it doesn’t specify when that fuel will be replenished. Also, for one month, gasoline is free and replenished in larger amounts. If you want to know when the fuel will be replenished, please contact the gas station.”
What we were actively discussing was why Dots has limits despite an OpenAI executive saying “virtually unlimited at the moment,” and whether those limits vary depending on the account or the task
What am I supposed to do now? Or do you mean, “Please don’t discuss the limits imposed on us any further in public chats”?
I recreated my dot and ran it with six agents, but I hit a limit after about eight hours. Looking at the WebSocket messages, it doesn’t seem to be a preview limit, and 29% of the context window is still available. If this isn’t a preview limit, why am I being blocked for such a long time as an abuse prevention measure?
My hypothesis is that there’s a bug in how these usage limits are calculated, causing my available quota to decrease every time I recreate Dots
I suspect there’s a bug in how rate limits are handled when recreating Dots
My theory is that if I recreate a Dot after partially consuming my usage allowance, the remaining allowance at that moment becomes the new maximum limit
Yeah, it seems Dots has a daily limit of roughly 3 billion tokens, including cached tokens, though this isn’t officially confirmed…
