Gemini thinking models not outputting to text selection
Just installed Windows v7.1. I'm running into an odd issue where Gemini thinking models such as Gemini 2.0 Flash Thinking refuse to output to my text selection no matter the program when running 'Rewrite' or 'Proofread' tasks. I've tried in browser (chromium) and in notepad. Gemini 2.0 flash (non-thinking) is working fine though. Gemini 2.0 Flash thinking is able to complete the tasks which output to the writing tools popup gui windows like 'Table' and 'Concise' tasks. There are no visible error logs which pop up on running the rewrite or proofread tasks. I know the thinking models are running given that 'Table' and 'Concise' work. Shortcut: ctrl+`
Not sure what's happening here as I would expect an issue open for this already given how consistent the behavior is on my end.
Thanks for bringing this up - the thinking models used to work fine haha. I can't test them at the moment on my end, but I'll take a look as soon as I can. Google probably changed something with their API, especially considering their new thinking budget parameters and such for the Gemini 2.5 Flash model.
Like you said, it looks like there is an issue or change with the thinking model. They now state "Gemini 2.5 Pro and 2.5 Flash come with thinking on by default. If you're migrating from a non-thinking model such as 2.0 Pro or Flash, we recommend you to review the Thinking guide first."
Looking forward to 2.5 integration.