Don’t worry guys, they know its so bad and worse.
We will see 2.5 Pro Experimental again as “new brand” model, believe me.
2 Likes
It’s still capable. You just have to be ultra-specific in your instructions. It’s not as intuitive as the March model.
1 Like
I’m getting good results again… I haven’t modified anything. But just now, both Gemini Flash/Pro preview are working fine. Anyone else experiencing an improvement? Maybe they have updated something due to Google I/O today?
Hi : has anyone tried calling gemini-2-5-pro-preview-05-06 using global or vertex api from aws lambda and what is the response time compared to normal machine or using google ai studio?
2-3 hours of “trouble shooting” why the right hand PREVIEW screen went totally blank (solid black) - I was amused, thought I would see if it could figure out the problem with it’s own code - LOTS of LLM “Modified by moderator”… in the end, it NEVER could figure out the error in it’s own code. Complete waste of time
1 Like
Same here. I find it really hard to get the model to generate code that is actually complete with the updates requested, and doesn’t come with a giant bunch of unnecessary and annoying code comments. As well as ocassional simply ignored sections of the code; very unusable.
2 Likes