Good day all
We are launching our new product and need to vectorize a lot of data. We’ve chosen to use the small vectorizing model. Unfortunately, we are hitting the quite low vectorizing limit and this is causing a bottleneck for our org.
Is there any way to get an exception? Otherwise we might be looking at a 2 month delay
Thank you
D
You’ve got $0.10 per minute in spending capped by the rate limiter (or less depending on the estimated tokens), or perhaps $100 per day if you are extremely persistent. In comparison to other models (like Sora 2 where you can spend hundreds per minute), it does seem paltry.
The batches API is a separate pool of rate limit, and adds 100 “minutes worth” to that. Then, if the batch jobs are done quickly and exit the pending queue, that should mean you can achieve more than 500M tokens per 24h promise. You can see that tier-5 is a significant jump in batch queue - still $40/max/queue after discount (over the $5 ensured additional spend you might have per day otherwise in batch with tier-4.)
Thus, I extrapolate that your total desired spend on 3-small embeddings is $6000 if you and I have done our math right. You can pay the additional needed to reach $1000 total if over 30 days since the first payment and graduate to tier 5.
There is no more “request exception” in the platform site any more. You’d have to go to help.openai.com, ensuring login there against your organization owner account, and send a message that explains your need to send OpenAI a lot of money for upfront AI usage to a particular model.
sales@openai.com is another avenue, but you’ll likely only hear back if they prequalify your interest and a Googling as “enterprise partner” material. Also Microsoft Azure services.
Note: this is a significant investment that will only work against a hosted model now two years old…
