"Slow" requests wait times and failures

Minor rant… I get that Team Cursor is presumably still working on the pricing model as things continue to grow. But if you’re going to do the whole fast/slow requests thing (especially the 10-long per day part) you really gotta have it work every time.

I’m paying for 1000 fast req/mo on Pro. I don’t make any money with what I code, so it’s all out of my own pocket. I see the value in it, but it’s not a cheap proposition and I really have to budget for Cursor use.

So I’m sure you can imagine the frustration of asking a question that the AI gets wrong several times in a row, re-asking without a satisfactory reply and hitting the entire limit for the day without resolving something adequately. Not to worry, slow queue right? Within a few attempts, 320 seconds wait for the message to be processed… watching the countdown in pain… and then getting “Connection Failed”. Retrying, and being told you have to wait another 320 seconds for the next roll of the dice.

That doesn’t feel very pro. It certainly doesn’t motivate me to flip the switch to be charged extra for the chance of more failed requests.

Related: Claude-3-opus slow requests with 320 second wait 100% always connection failed (This isn’t just for Opus though).

I’m sure this message will also disappear into the noise, there seems to be a lot of community posting on here of late but vanishingly few staff replies at the moment. That’s not great either. I’m sure you’re very busy, but still. If you do at least read this, please take these kinds of negative experiences into account when you’re figuring out your charging structure and designing the user experience. Thanks.

3 Likes

Hi @three

This is unfortunate. I hope the developers will resolve this issue soon.

2 Likes

yeah I’ll bump this. This is insane. One positive thing out of this though is it certainly is motivating me to learn code so I can stop throwing my money away with Cursor.

2 Likes

300+ now

Edit: 350…
Please add servers above >100 :slight_smile:

2 Likes

A
Mayday!

1 Like

Niceee, let’s see who can top position 407. Anyone? Anyone?

2 Likes

Actually, I suppose this might top it actually.

and then

3 times in a row now! :wink:

1 Like

Screenshot 2024-11-18 at 3.50.23 PM

Didn’t cursor get like 80 mil in venture funding a few months ago? Where is it going lol

2 Likes

I’ll bump this too. This is insane.

1 Like

bumping this up, seems this is a huge issue currently

1 Like

sam here too . slow requests became unbearable

1 Like

I really agree here! I’m paying per month for Pro, and I lost a whole workday yesterday, because I didn’t even get the slow request through!
I wait and wait, and then it is Cencelled with “Slow pool too large!”. This is not acceptable. Cursor is getting payed for things they do not manage to deliver. They need to fix this ASAP, or me and many others will cancel their subscriptions!

1 Like

I’ve noticed a dramatic change in my API request consumption. Previously, 1000 requests would last me nearly a month even with heavy usage. However, after switching to composer in Agent mode, I’m burning through the same 1000 requests in just 5 days.

If others are experiencing similar patterns, it would explain the increased load on the Slow Pool. I’m hesitant to upgrade to a $160/month plan to cover these needs, so I’ve unfortunately exhausted my quota this month. For next month, I’m planning to reduce my agent usage significantly.

Same here! This is not OK for paying Pro-customers.

1 Like

Likewise I’ve hit my limit for the month and it hurts. It’s heartbreaking to wait so long and get either of:

  • Slow pool too large
  • Anthropic rate limit hit

Shouldn’t it be possible to show “slow pool too large” when the request is made, rather than delivering it after the long wait?

2 Likes