What’s the point of Request Per Day for new experimental model ? Anyway it will be turned off in very short time. How i should test it?
Seems like it have the same RPD as Gemini Pro - x50.
1 Like
Welcome to the forum.
The rate limit is 50 rpd based on experiment (I hit the day limit). The model is clearly Pro class, not flash family, that’s obvious from the quality of responses.
First day of testing, it had a high likelihood of running into internal server error; probably too few instances of the model in the server farm and too many arriving requests. Most requests had to be repeated on average 4 times to have one request get through.
You’ve reached your rate limit. Please try again later.
I think it’s not because of the high workload.
1 Like
Does anyone know what the limits on tokens are?
The limits are so damn annoying.
50 RPD??? Wow, that’s rough. After some initial good results in legal contract analysis, I tried using the exp-1114 model in a workflow I’m building. Not only is the RPD low, but RPM/RPS must also be throttled down. My pipeline makes 6 small requests then combines them into a single answer (pseudo-map reduce to review longer docs), and I was getting 429 errors.
I’m excited to try out new model snapshots, but not if I need to throttle and rewrite everything to test it. Also, can’t run a basic test suite with 50 RPD. Need at least 10x to get any kind of evaluation done.
1 Like