Hi, appreciate the diversity in available model options to choose from. I’ve had a good experience using flash thinking.
However, the output formatting in the chat window is not consistent and seems very off many times, which renders it difficult to use. Was wondering if this could be fixed soon?
For example, when the model is outputting code, unlike other models like 3.5 sonnet, the code is not clearly in code blocks that can easily be applied to my file. It sometimes spits it out in a format that’s impossible to read.
yes i was gonna post about it too. gemini 2.0 flash and its thinking model deserve more attention from cursor team. it is pretty ■■■■ good when it works. non thinking model works sometimes, thinking model never outputs the markdown correctly, so there must be some bugs.
after they get fixed, people will know how capable they are, and that can unload some traffic from claude 3.5 sonnet.
I’m actually getting random format issue too , not just for flash i think it depends on what you’re doing , might need to adjust your prompt abit , sometimes i even see the isntructions set by cursor team LOL. Anyway just report them and let cursor team fix it
please fix, i alrdy reported this 5 days ago. 2.0 is such a great and quick model, its really unfortunate when it cannot properly output formatted code that i can apply to my project.
Hey, Flash 2.0 is still somewhat experimental, and we don’t believe it’s a big enough improvement to warrant the time and engineering work it would take to improve the performance right now, but we definitely haven’t ruled this out for the future, especially if Gemini models get progressively better / cheaper!