I just got usage capped on GPT-4 after 20 messages – when I clicked “Learn More” on the message, I saw:
Thanks for your interest in GPT-4!
To give every Plus user a chance to try the model, we’re currently dynamically adjusting usage caps for GPT-4 as we learn more about demand and system performance.
We’re also actively exploring ways for ChatGPT Plus subscribers to use GPT-4 in a less constrained manner; this may be in the form of a new subscription level for higher-level GPT-4 usage, or something else.
Please fill out this form if you’d like to stay posted.
Now admittedly I paste massive chunks of code into GPT-4 as part of my daily workflow and it’s understandable if they’re wanting to make the amount users get match with the price they’re paying… but I was still a little taken aback by the customer-facing bullshit of that whole “To give every Plus user a chance to try the model” and “as we learn more”. Like bro if you feel like setting a limit based on use then just tell me what the limit is and how I can get more if I need it.
Anyone else run into this? Anyone have a good alternative (besides just sending it all to the platform API and paying out the ass)? GPT-4 is actually capable with code in my experience in a way that 3.5 and Copilot are not.
With the API I’m paying less than 10% of the subscription fee.
Just how massive are we talking about?
Don’t you need a subscription to use the gpt-4 API?
Last I checked, gpt-4 is 0.02 for 1000 tokens. Every message in chat also has a summary of the whole convo plus the most recent messages. I feel like that’s busting the 10% pretty quickly if it’s intensive daily use.
The tokens don’t have a fixed price, the 2 cents are an average depending on the complexity. I’m using it moderately almost every day, and have rarely paid more than $2/month.
And no subscription needed, it’s prepaid.
You need a subscription either way
GPT-4 costs from 0.01 to 0.12 per 1000 tokens depending on some details – but regardless of that, it’s not like chat type chat where you might have tons of small messages which each depend on the full 32k or whatever of context; each singular message usually has an explicit context for the stuff you want to tell it, and no more than 50-100 of them per day to implement your thing at most, so like 50 cents to a few dollars a day even at an obscene level of usage. Might be more than $20/month in total but more likely less.
You don’t need any subscriptions to use the API. It’s pay-as-you-go.
Oh, I was meaning in terms of “you have to pay for it” – yes, you’re 100% right, you don’t have to be on the $20/month thing in order to use GPT-4 on the API.
You don’t need a subscription. You just buy credits and it’s pay-as-you-go.
Source: me as that’s how I’m using it. No subscription fee / recurring payment needed.
Do you use DALLE via API?
TypingMind w/lifetime license works beautifully for cheap simultaneous GPT-4 Turbo & Claude-3-Opus when it comes to text. And can upload images. Generating would be interesting, don’t believe it can do it.
Nope I’m only interested in the word processor / coding bit, so GPT 4 is all I need. I’m accessing it through https://github.com/Bin-Huang/chatbox.
So you just set up your own interface and then make requests there? I did set up a MERN stack app for chatting with it as an experiment, but I never did anything else with it after that.
Not even that, I’m using chatbox where you can simply add the API code in the settings and be done with it.
The software integrates a bunch of other AI’s, some of them in Chinese, but I’ve removed most from the quick access menu and only really work with GPT.
This morning was 177kb in and out, so call it 2/3 of it is input and 1/3 output, would mean roughly:
118k bytes input ≈ 29k tokens = 29 cents 59k bytes output ≈ 15k tokens = 45 cents
I think you may be correct in your assessment
So you pay for it and you received that message? I pay for it and paste hundreds of lines of code into it, making request after request for hours at a time and have never received that message. Maybe the system was overloaded? I’ve found that when I try using it through the app it will error out sometimes, and take a long time other times, but I’ve never encountered that on a desktop.
I unsubscribed because they blasted me with 5 min “are you a robot” tests for every request.
Now I’m using Perplexity, that sucks too but at least it is usable.
Don’t forget “we have detected suspicious activity coming from your system” and “there was an error constructing a response” (or whatever the wording is) which form roughly one-third and one-third respectively of the responses I get from it, along with the one-third that are successful responses.
