While gpt4 provides some better solutions than gpt3.5, gpt4 still makes a lot of mistakes due to its knowledge of APIs and SDKs changes being out of date. As a result, it can take multiple corrections for it to provide a working answer. So, a limit of 40 would make gpt4 a lot more usable than the current 25. I agree there should be a cap, but due to the number of mistakes it makes, 25 really isn’t enough.
Another way gpt4 is better than gpt3.5 is that it doesn’t try to re-print the entire code block each answer, only to repeatedly fail to do so entirely. The continue seems to work better in 4 than 3.5, so fixing the continue in 3.5 would be a great help. The biggest issue is with code block continues in 3.5, which will continue the code in plain text and the explanation in a code block. Sometimes I can get it to continue the code block in a code block, but then I don’t get the explanation appended.
It’s a great tool and worth the subscription, but its limitations can be pretty frustrating to deal with.
I echo the concerns already raised in this thread.
GPT-4 is a genuinely useful tool for a wide variety of technical applications, but the 25 message limit is absurd, and means it can effectively only help you with one or maybe two issues every 3 hours.
If GPT-4 really is just that computationally expensive, then change the pricing model. 25 messages is far too limiting for how expressively useful this tool can be.
The current message limit wastes enormous human cognitive effort in battling yourself back and forth:
Is this question worth asking GPT4?
Let me be very careful asking GPT4
Should I experiment with a new way of asking my question? No the limit will get me
And so on and so forth… GPT4 is impressive but the limit needs to be more like 20 per hour for it to be useful or worth it.
Furthermore the limit should have some sort of rolling mechanism. Someone who asks 25 times to get a solution in 3 hours versus someone who uses all 25 asks every 3 hours are paying the same amount. This pricing model rewards workaholics and punishes everyone else.
I agree the cap limit of 20 isn’t very practical, especially knowing the cap may get worse. In today’s modern world of scalability and the amount of money being pumped into Open AI, coupled with the subscription we’re paying, I think the amount this cap is set to is questionable.
The other thing which is quite frustrating is that code blocks it writes stops randomly. Which means that you’re forced to use more to get it to keep going. If you’re going to put a cap on it, please make it so premature cap blocks and our responses to get it to keep going isn’t counted in the cap block.
It is impractical to limit the window for 25 prompts within a 3-hour timeframe. Additionally, once you have experienced the increased capabilities of GPT-4, it becomes challenging to revert to the default GPT-3.5.
True. After the lapse of the 3 hour cap, there should be an option to revert back to GPT-4, If one decided to continue a chat using the 3.5 when the cap was reached
I’m getting extremely frustrated by the 25 limit. I’ve tried reverting back to 3.5 with one conversation but was completely disappointed with the way it was handled and I couldn’t convert the conversation back to 4 again, so not something I will ever do again!
I’m rather disappointed with the subscription as I use this for various purposes from work, personal projects, a study aid. 25 messages on 4 is too little and I find that 3.5 just doesn’t work well enough for me to make it worth even asking at times.
Do I need to pay 20$ for 25 messages?! It is not a small amount of money for me! It is really unfair and very frustrating! You actually made people believe they can use GPT-4 freerly with no limits! And now you put a limit with 25 messages?! It is unfair and it is a manipulating way to make people pay money for limited amount of messages!
I would like to see an update on this. If there is going to be a usage cap indefinitely there should be additional subscription tiers to bypass or at least increase the cap.
I totally agree with the 25 limits being rediculous. I only recently back to creating apps and programming after decades. I find that after the 25 message limit and reverting back to 3.5 that the expereince is akin to going from Top tier support to level 1 support.
Just signed up so I can use GPT4, just to find out it’s limited to 25 messages per 3 hours. Haven’t typed a single message yet, but already regret getting this subscription. I use GPT for coding and 3.5 tends to “forget” steps along the way, so I need to tell it over and over what I’m trying to get to which is very frustrating and time-consuming.
Hey folks, thanks for all the feedback. The reality is that as we scale our services, we need limits in place. There is not enough capacity to give everyone the amount of bandwidth they want right now, and we are aware of this.
Thank you for being patient as the team works to scale with the unprecedented demand! I am going to close this thread for now, but please note, we hear you, and are working to improve ChatGPT across the board.
