Ask HN: OpenAI Elevated Error Rate?

1 pointmartinald4 comments

For the past few days, and especially yesterday and today I'm getting loads of failures with 429 errors saying either server error or the model is overloaded, on gpt3.5-turbo via the API. There's not much on the status page. Are others experiencing this?

View on HN

It happens all the time, it's not a good idea to build any serious products on their API, we are still early in the AI offerings, there will be better alternatives, I would look to the open source models you can host yourself

Yes definitely but the broader point is the lack of transparency on status pages. It's imo got vastly worse than it has has ever been in the past few days.

FYI, 429 is too many requests, that is your error not theirs

No it's not. The response body says they are overloaded, I thought that at first until I noticed the response body. We are also nowhere near the API limit (they return this in the headers).

Reply on news.ycombinator.com