Why's GPT 4o insanely limited to free users and even plus users? It literally barely gives you 5 messages in 5-6 hours to the free users

Your unlikely to find any service which offers unlimited access to frontier models at a consumer-friendly price.

The fact is that every message exchanged results in some cost to the provider—not to mention the costs associated with developing the models to begin with.

A year and a half ago no one had any access to generative AI anywhere near this capable at any price and now people are complaining about the fact they’re limited to 1 message every 135 seconds for $20/month.

@elmstedt The price is fine, it’s the service that’s getting shaky. A per message cap on paid users is just a big ol’ middle finger. Especially with how the AI behaves currently conversation-wise.

“A year and a half ago no one had…” yes, yes and I shouldn’t be mad when I stub my toe because there are starving kids somewhere. I don’t mean to be rude, but that’s simply irrelevant. I’m talking about the fact that the service went from free to paid, ok cool, then went from great to good, good to okay, and now poses limits on me even if I pay. That’s the part where I feel taken advantage of.

If you can’t understand why it would upset so many users, I think it’s highly likely we just use it in different ways for different purposes. Using it as a coding assistant, for example, is less like having a master programmer and more like having a newbie intern. You will often be pointing out mistakes and correcting the AI. Sometimes it even responds in the wrong coding language like giving you a page of Python when you’re working in a strictly JavaScript environment. Like I said, a per-message cap feels like a big ol’ middle finger. At least let me pay another 3 to 5 a month to get rid of it or something.

I also don’t sit there and use it every day. Sometimes I pay for a whole month and only need to use it one time that month, but hard and all in a 6 hour window. So many reasons this is lame.

Personally, I have both a Plus and a Teams account but in practice I only very rarely ran into message cap issues even with the 40-messages/3-hours with GPT-4.

With GPT-4o I’m not sure what I’d need to be doing in order to burn through 80-messages in 180-minutes.

Between the time writing my messages, adding relevant context, the model writing its message, reading the message, and taking action on the message, I just don’t see myself doing that non-stop for three-hours straight.

If the message caps are so onerous to you, which really seems like that’s the issue given what you’ve written, maybe switching to a Teams account or running a local model instead would be better for your use case.

I thought the “Teams” version also had message limits? When I looked I didn’t see any option or plan that got rid of message limits I think they just raised the cap a bit. If the “Teams” version gets rid of limits I’ll definitely look into that.

I too only rarely hit the cap as it is but when I do, boy is it inconvenient. A couple days ago I needed to run 1000+ lines of x,y,z coordinate rewrites that involved some heavy math and pasting it in big chunks was wrecking performance so I cut it up into smaller bites like “okay now do this one”. That was just 1 of the handful of tasks I was expecting to complete that day. I had just spent 20 on a month I didn’t even use then paid another 20 to reactivate specifically for that day’s tasks and within 1 hour 30 minutes I was locked out and told to use the other GPT (which had absolutely no context to the conversation or workflow) and had to start over but by the time it was rolling smooth I got locked out of that one too. I’m lucky that particular day’s work was only for a hobby. I’m horrified to think of that happening with anything important.

I haven’t looked into a local model yet, I’ll check that out too. Thanks, elmstedt.

1 Like

I use 4o as an interactive study tool preparing for my medical licensing board exams. I rip through the limits as a paid user as well. It’s so damn useful for building a database of information and actively quizzing yourself on that information, expanding on said info (if need be). I just wish there was an option to purchase a higher tier without hitting limits. Ive loosely gathered that using the API version and purchasing tokens may allow for this… however I’m absolutely ignorant with regard to programming/ coding. If I had to guess what an API is, I would say… a government agency? Obviously not, but if this is a possibility would this option be way too far over my head to pursue? Would I even be able to achieve higher limits? Or am I doomed to the limitations. Thanks to anyone for insight.

The enterprise version has no limits, it starts at $108,000 if you can get some friends to help you it might suit you .

googling discount/promo codes rn. meanwhile, do you know the limitations for Teams version?

There are no discount or promo codes. From the people who I have spoken with who use Teams, they would guess about 150-160 messages every 3 hours, but have never hit the limit.

I meant to quote reply the message above regarding the $108k for enterprise and jokingly say I was searching for a promo code. I did not actually expect it. But thank you for the information!

Nothing encourages users from trying out other AI tools more than Chatgpt unexpectedly crashing or displaying an error message in the middle of an important task. :hand_with_index_finger_and_thumb_crossed:

I just got this and I’m deeply disturbed. How can it be per day now???

Is your account free or paid?

paid. i’ve had it since it they offer the plus.

Thanks… Looking closer at the message, I’m wondering if your post isn’t a bit unintentionally misleading?

Correct me if I am wrong, but it appears this message is specifically in response to a request for a Dall-E 3 generated image, yes?

This is a little different than the topic being discussed which is the 3-hour window message limit. Image generation has always had different limits than message exchanges, including a daily limit on the number of images which can be generated.

Here is a topic from 8-months ago discussing the daily limit for image generation in ChatGPT,

The rationale for the separate limit for image generation is likely that the Dall-E 3 model is much more computationally expensive than an average chat message generation.

So, for better or worse, this is where we are. Image generations have a daily cap and have since at least late last year.

1 Like

I have a free and a plus account, and I was testing my free one, and it let me use two messages of custom GPTs before it said to wait another 5 hours for another 5 messages.

This is insane. I mean I also have a plus account and I was so close to cancelling it when the announcement of GPT-4o coming to the free tier. I’m so glad I didn’t. GPTs, and GPT-4 is so limited in the free tier it’s impossible to use.

Upgrade to Plus if you actually need to use GPT-4. Don’t rely on the free version.

Hey, we have some automated bots that flag posts so your post was probably automatically taken down, it is now back up meaning a moderator restored it.

1 Like

Def not an expert here but here’s what I think/do. New tech takes a moment to catch up (yeah I know they probably have had time). The method they use to balance compute probably isn’t the best but that is cutting edge stuff. In the mean time while you are waiting if you need a quick response or query from open AI consider using playground and paying for your actual compute time with an API key. This is what us devs do. If you need actual commercial considerations this is your path. I like my simple gpt’s I’ve made too but when I’m not farting around and need real data this is the way. lol

I don’t like dealing with shortages.

Dear Jake Elmstedt, Since you are a leader, I guess you are a contributor from OpenAI. Therefore I would like to submit a suggestion:
What about having a counter of remaining prompts displayed at the top right of the page or anywhere visible so that we as users adapt the usage and do not get stuck in professional use?
When the quota is totally used, I appreciate knowing when I can come back to ChatGPT to continue my work. Here is a message I got not very helpful in continuing my job: “You have reached the maximum number of prompts you can send per day.”
I am a plus user.