I’m trying to figure out what’s going on with Gemini 3 Pro, because for the work I do it has been a big step backward from 2.5. My usual workflow is to upload large batches of scanned research material—often 40–100 PDF pages of mixed notes and documents—and have a custom Gem turn them into a clean report with added context. In 2.5, this worked beautifully. It could digest huge amounts of messy archival material, pull out the key points, and fill in missing details with accurate online searches.
Since the upgrade to 3 Pro, the same Gem has started inventing information, ignoring instructions, and failing at simple tasks it used to handle. For example, it now claims it “can’t find” basic author life dates that 2.5 located easily. I can’t find any area where 3 Pro performs better for this kind of long-document work.
I’m trying to figure out whether I need to rebuild my Gem from scratch, whether the model needs some sort of adjustment period, or whether others are seeing the same issues. I feel like I'm losing my mind over here, because everyone is seeing a kind of revolutionary progress but I know that it's for very very very different kinds of tasks.
Thanks for your thoughts, everyone!
P.S. Is there any way to still access 2.5 Pro as a regular Pro user?