Flash 3.5 Changed, Downgraded?

I’ve been testing a workflow over the past 2 weeks, across different providers, models. Flash 3.5 was the clear winner and very consistent.

Something changed over the weekend. No change to my prompts, code. Workflow has multiple agents.

Agent 1 that was completing in 3-6 seconds, now takes ~20-30+ seconds.

Agent 2 was completing in 20-40 seconds. Now takes minutes. This is also a grounding agent. The searches it does also exploded from ~10-20 to 30-40+.

Anyone else seeing a change? (changing thinking level does not change the outcomes, slow across the board).

Thoughts?

Hi
Can you share some mored details with me

  1. Sample prompts , Sample outputs ?

You can DM me these

DM sent. Thank you kindly!

I noticed it yesterday. I don’t mind that it started working very slowly, but it went into a “loop” and by the time I figured out what was happening, it had “eaten up” my entire weekly credit. Of course, nothing was done, and the task I gave was just a check and correction. There is a lot of discussion on the Internet about the fact that AI should only be paid for work done. Yes, it works much worse than before. I don’t believe that my weekly limit will be restored, although I’m sure that this problem has already been noticed by support.

Update : Yesterday I found this issue resolved , so thank you ! :slight_smile: