Google just dropped Gemini 3 Flash! It’s available in Cursor now.
Have you tried it out?
In the announcement it shows it has 1M context window, in Cursor it shows 200k context window, so which is it?
Also what about pricing? No free week?
1 Like
Hey @DavidWasHere!
If you enable Max Mode, you can use up to 1M context!
Promotional pricing isn’t guaranteed for each new model release. It depends on the model provider. There’s no promotional period for Gemini 3 Flash at this time.
2 Likes
Was Gemini 3 Flash added to the Auto pool since it’s super cheap? I just opened a new chat and used Auto and it was blazing fast as compared to my previous Auto chats
“I tried it, and so far it was much better than Composer 1, and in my opinion also better than Opus 4.5. It was definitely better than Sonnet 4.5.”
This model is excellent. It is affordable and supports Cursor’s plan mode. Unlike Haiku or Grok, Gemini Flash 3 handles planning tasks well. It finds bugs and implements new features easily. I really enjoy using it.
1 Like
It’s a great model, no doubt—but could you please tell us when you’re finally going to squash the “planning next move” bug that even my grandpa has heard about?
3 Likes
I’m frequently getting 429 errors with the new Gemini 3 Flash model when using an API key
No, because “You’ve hit your usage limit”…
I find it weird how a fast, efficient model like this is also limited, so we have to spam retry on Auto mode to get a decent model to solve problems.
2.5 flash was unlimited.
1 Like
same here
This Model is Great, Thanks for add to cursor.
but please add medium reasoning of this model.
1 Like
This model, according to google docs, should be very good at reading pdf files and other stylesheets. Has anybody tried this? I am not sure if this is in any way limited by Cursor’s harness, but in my experience it is basically unusable for this scenario.
1 Like
perhaps since we use in Cursor as IDE. try in Gemini Chat, its work as expected
1 Like
I did, it is indeed better, I am not sure if this is limitation of Cursor or if Google does not allow this kind of features through their API, but it would be really good if I could do this directly within my IDE. In my work we get a lot of reports in PDFs and I have to extract all relevant information manually
1 Like
ask cursor to create python script to extract all relevant information? Why not do that?
1 Like