Response: “Rate limit reached for gpt-4o-mini-realtime in organization org-xx on requests per day (RPD): Limit 1000, Used 1000, Requested 1. Please try again in 1m26.4s. Visit xxxx to learn more.”
How to query the rate-limit value of the gpt-realtime-mini model in real time. The example shows that the ‘rate_limits.updated’ event can provide information on the consumption of requests and tokens. However, in reality, it can only return the token consumption, but not the request consumption.
Additional:
Q1:
I was using gpt-realtime-mini model. Why did it show “gpt-4o-mini-realtime” in the response?
Q2：
Does RPD refer to a single natural day (i.e., automatically refreshed at 0 o’clock) or to the past 24 hours?
Q3:
I actually have a large number of requests every day. How can I increase the upper limit of my RPD? After Tier’s upgrade, I noticed improvements in RPM and TPM, but RPD did not show any increase.
Q4:
Are the values of gpt-realtime-mini and gpt-realtime shared or independent? Also, what is gpt-realtime-1.5 ?Is it an upgraded version of GPT-Realtime or does it have slight differences in usage from GPT-Realtime?
OpenAI removed the page in documentation that had all models with each limit laid out, selectable by tier.
Thus, you cannot see the results that your tier should be at for tokens-per-day, as it is not shown in the models page, nor see what you would get if paying even more to go to a higher limit.
For example, to find out that a organization has a cap of 160,000 TOKENS (not requests) per day to gpt-3.5-turbo-instruct, you have to go to your own “limits” page for the organization.
At tier-5 though, I can report, there is no daily limit of requests for gpt-realtime, just tokens. You can see yours here.
The headers should return values with the current rate limit, current requests remaining. However, it doesn’t have a separate field for “per day”, and doesn’t say that this would be returning the minimum of either remaining requests per minute or requests per day.
Other rate limits are “windowed”, as if they refill continuously at your rate by the algorithm. You can see in your own API error message, you are told to retry in 1.5 minutes, which is indeed a “refill rate” of 1000/1440 minutes per day.
I’m currently at Tier 2. I can’t see the RPD limit of the gpt-realtime model. did “-” mean the number of inheritations from the previous level or only an unlimited number of times
If you are getting limited onh requests per day, yet in the org rate limits page, there is no RPD for that model in the right column of the limits page, then it would seem that there’s either a “shadow” limit, or the organization is limited incorrectly.
You should contact thru the help.openai.com bot, firmly, with your need to have the wrong limit removed.