Gemini 3 Pro with Nano Banana — “model currently overloaded” in most generation attempts

Hi ​@Eugene_Simonov, 

I understand that you’re experiencing the “model currently overloaded, try again later.” error when using Gemini 3 Pro.

Thanks for flagging this. I’ve passed this along to our team internally.

In the meantime, while you mentioned that the results may not fully meet your expectations, you could try switching to Gemini 2.5 or GPT 1.5 as an alternative.

Thanks,

Thank you ​@Junko3  for your quick response!

Let me know if you need any additional information from me. 

Yes, of course, I am now trying to use Gemini 2.5 or GPT 1.5 more often, but of course, I would like Gemini 3 to work as before, because the results were often very successful.

In any case, thank you for bringing this to attention and trying to resolve the issue.

I wanted to report that I am facing the exact same problem described here. I encounter these overload errors with both Gemini 2.5 and 3.0 models. Switching to other models isn't a viable solution for me as they don't fit my specific workflow requirements. My colleagues are also affected by this same issue. 

This problem has been going on for a long time. I have a Google subscription, but I find it easier to work in Figma. Is it possible to implement the ability to use nanabanana's own APIs?

Hey everyone! I have same problem

Hey All, thank you for continuing to flag this issue! 

Our team is aware of this ongoing issue with Gemini 3 Pro with Nano Banana. To be completely transparent, there has been a significant number of reliability issues with it this month, and it’s been determined as a Google issue that we cannot control. As a temporary solution, the team encourages you to use Gemini 2.5 or GPT 1.5 instead. 

We’ll continue to pass your feedback onto the team and will provide any updates addressing the Gemini 3 Pro with Nano Banana issues, when/if we hear more directly from the team. 

 

Yes, I have the same problem. Now Gemini 3 Pro hasn't been working at all for several days.

Hi ​@Aldrei and ​@Eugene_Simonov, thanks for following up! 

I completely understand how this must be affecting your workflow, but as I mentioned above, this is a Google issue with Gemini 3 Pro with Nano Banana that our team cannot control at this point, which is why it is now unavailable for use. As a temporary solution, please try using Gemini 2.5 or GPT 1.5 instead. 

Hope this helps clarify. I’ll be sure to provide an update from the team as soon as we know more about the ongoing Google issue.

Hi,

Thank you for the explanation regarding the Google issue.

However, this raises a serious concern. If the feature was known to be problematic or unavailable, why were credits still being deducted for failed attempts? I was not aware of this issue, so I kept trying to generate images, resulting in all my credits being wasted without a single successful generation.

Furthermore, I tried the alternative models you suggested (Gemini 2.5 and GPT 1.5), but they are not working either — they also deduct credits but fail to produce any images.

How am I supposed to get these credits back now? What solution do you propose for this problem?

Hi ​@Web-designer CS, thanks for following up about this issue! 

I flagged this over to our AI team, and they confirmed that failing image generation attempts do not deduct credits. They’re now taking a closer look at your account to investigate what’s happening, since deducting credits for failed attempts is unexpected behavior. 

I have the exact same issue. Figma needs to fix this. We are paying customers, and when you offer a service, and we PAY for it, we expect it to work. its 2026. This is not OK. 

I have been using Figma for over 4 Years now.

I started to be a paying customer and AI features are one of my fundementals, for basic icon and other things. Obviously they cant do anything complex. But for the past 2 months

Any reasonable model I pick is  currently overloaded ERROR with any day and time of the day.

This is NOT OK. You are a billions dollars worth design software company and you cant have this for too long. 

People. Dont be OK with this. 

I demand all the credits and time wasted getting this error back, and expect things to work normal.

Hi ​@Gokhan Uzmez, thanks for chiming in here!
 

I hear how frustrating this has been, especially as a long-time user relying on these features. You’re absolutely right to expect reliability from paid features and repeated “model overloaded” errors aren’t the experience we want you to have. The team is actively working on improving capacity and overall stability.
 

To clarify, failed generation attempts (including “model overloaded” errors) do not consume credits. Credits are only used once a generation successfully completes, even if there were retries beforehand.

That said, I understand the bigger concern here is consistency. If you’re still running into this frequently across different times of day or models, that’s really helpful for the team. If you’re open to sharing a bit more detail (like which models you’re using and when you’re seeing this), I can make sure it gets passed along.
 

We appreciate you taking the time to share this feedback and for sticking with Figma.