I guess you mean the gpt-4-vision-preview model?
https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo
There is currently no other “vision” related product on OpenAI.
I did some testing the past days and the issue seems to be resolved. Anyone else still experience issues?
@nikunj is currently working on this issue. At least that what he wrote me.
Maybe he could tell us if there were changes implemented.
It was planned that he will comment on the initial post from me to keep us updated.
https://community.openai.com/t/assistants-api-gpt-3-5-turbo-16k-usage-exceeds-limit-due-to-message-loop
We haven’t implemented any changes yet. This requires improvements to our model behavior and we plan on making these improvements as part of our next model release early next year. Model behavior can unfortunately be unpredictable (i.e. things work well sometimes and not other times) so it’s possible that things are working better for some folks at the moment.
2 Likes
I wonder if all development resources are going toward the microsoft end of things and chatgpt will die out as a business. It seems that once the money started pouring in that it needed to go to a for profit model with no caps. I love the product but there are major issues with implementation and calculating costs which makes it impossible to wrap any other business models around that. I hope it’s just a temporary issue that gets resolved but its frustrating that they offer these things, charge for them and then the quality of the support is non-existent or very little with very little supporting documentation. I am looking into attacking this from the Microsoft angle as i think that ecosystem will be more mature and predictable. 
Thanks for your information.
Quality of AI models in my opinion should be OpenAI’s laser focus, in terms of quality of satisfying operations defined in their programming and user input, for scenarios that are not ChatGPT, in a manner already seen but fading, and warnings, denials, distrust of the developer, avoidance of mistruth, and benchmarks should take a back seat to meticulous attention-filled error-free 0-shot task inference. There is no $80B valuation without a competent AI.
This topic is full of varied concerns, possibly ranging from billing errors to obtuse assistants without controls, and has become rudderless.
1 Like
I experienced the same thing. Went jumped 10 fold.