I only got one gemini pro 3 (preview) prompt, same for flash.
2 Likes
Is this the end of AIstudio? I have a PRO account but gemini-app is totally worthless for testing and working with long context, already at 10k+ it’s valueless! AIstudio was a good place for testing, and Google was getting better training data than what’s in regular user-model exchanges. Today I checked and I’ve used PRO barely 10-15 times. This limit isn’t even suitable for checking 2-3 cosmetic changes in system instructions.
14 Likes
Yes, the RPD has been reduced—from 20 messages per day to 10—which is already lower than before. I hope this is only a temporary limitation before they release the stable version. However, based on past experiences, I’m concerned that the RPD may continue to decrease in the future.
8 Likes
if you want help here it is, sign up for google cloud and receive the $300 start up bonus, Switch your account to a paid account. then you have two months to use 300 dollars worth of token. cancel your paid account before your two months is up. And you wont have to pay anything. Or you can build something google wants and they will up your limit. They use your information to improve their products. I built a few agentic scaffolds and never hit a limit. I hope that will help you out.
In the two months I did not even use 20 dollars worth of tokens. 17 dollars is all it cost and I hit the million token a minute limit once an accident. But i was creating a Crisper cancer treatment So it was worth it. And I am giving it away if any one wants it?
1 Like
Please connect AI Studio to our Google One subscriptions. The rate limits are unbearable.
7 Likes
Definitely this. All the AI subscriptions and useage limits and stuff is so confusing the way it’s setup right now. I just want to use AI Studio to dev, Gemini App as an AI assistant that’s more intelligent.
4 Likes
Same google. You are DESTROYING your user base with this… Try different rate limiting? Maybe limit veo and genie more? But not Text queries? Getting 10 3.0 pro requests as a pro subscriber is LAUGHABLE… I hit that before 7:00AM… Back to Claude and ChatGPT for me…
6 Likes
You can try using Vertex instead. For some reason, the rate limits there are different from Google AI Studio. Even on the free tier, it usually gives you a higher requests-per-day (RPD) limit.
The downside is that you can’t save prompts on the free tier, and you need to manually select and set up the model first.
3 Likes
Hello, do you know approximately how many free daily requests Vertex offers? Not from official data, but just based on your experience. And do you know if I connect a billing account, will it charge for each request right away or will it start with a free quota?
I’ve been using vertex and the free tier for almost a year and haven’t been charged at all. If I remember correctly, the free limit was around 50 to 100 requests per day (RPD).
2 Likes
Thanks for the info. I’d also like to clarify what you meant by prompts not being saved on the free tier? Are they temporary chats or what? And if I have billing enabled, how do I know if I’ll be charged for chat requests or if the free requests expire first and then the paid ones start? I just can’t figure out how to enable/disable the API key for the chat in Vertex AI.
1 Like
About prompts not being saved (free tier):
In Vertex AI (especially when using Vertex AI Studio or the API directly), chats aren’t automatically saved the same way Google AI Studio saves conversation history. In many cases, once you refresh the page or start a new session, the previous prompts aren’t stored as a persistent “chat history.” So it’s more like temporary sessions unless you manually save them somewhere.
About billing and free usage:
If billing is enabled, Google usually applies any available free quota first. Once you exceed the free quota (like free RPD or free tokens), then paid usage starts automatically. There isn’t a manual “switch” between free and paid, it just depends on whether you’ve passed the free limits.
To check:
Go to Google Cloud Console → Billing → Reports to monitor charges.
Go to IAM & Admin → Quotas to see your usage limits.
You can also set a budget + alert in Billing to prevent surprises.
2 Likes
Thank you so much for such a detailed answer. Do you have any tips on choosing a region? Does it affect the limits? As far as I understand, you can specify several at once. Also, in your opinion, is there any difference in the quality of the models in Vertex AI and Google AI Studio? Or does it seem like they both provide access to the same model and performance? For example, between the Gemini 3.0 Pro, in the Gemini app and Google AI Studio, there’s a significant difference in quality in favor of the latter, even when using seemingly identically named models (3.0 Pro).
2 Likes
I recommend using the global endpoint, but the best option can still depend on your region and latency.
As for the difference between Google AI Studio and Vertex AI models: the main distinction is the underlying model version and data freshness. Vertex AI may use a slightly older base model (for example, trained up to late 2024), while Google AI Studio may provide access to newer model updates (early 2025 versions). But don’t worry, overall there’s no big difference I noticed.
That said, in Vertex you can enable grounding (such as web grounding) to get more up-to-date and accurate responses when needed.
Also, keep in mind that responses in Vertex can sometimes be a bit less restricted depending on configuration and safety settings, so it’s important to review and properly configure safety filters based on your use case.
1 Like
I have another important question. Am I correct in understanding that in Vertex AI, you can’t edit messages sent by a model, unlike in Google AI Studio? Is there anything that can be done about this, and are there any workarounds? In Google AI Studio, for example, you could also delete a model’s thoughts separately without affecting the response, but here, as far as I understand, both are deleted simultaneously? Also, is it possible to create branches?
Anyone using it for image generation… Are you finding it takes too many tries to get one correct image? My prompts are extremely detailed and I add photo ref for faces and footwear (since it seems to struggle with just text description) and it seems hit or miss. I’m also finding any of my characters with skin color darker than lily white get flattened, white-washed and the ethnic features homogenized. I’m using free tier because I can’t see paying money to get this kind of result. I seem to be getting 10-15 prompts and about 6-7 images. (PS– Gemini is soooo much worse for me. It should stick to picking out eye shadow for Lindsay Lohan!)
1 Like
Same here, Steve. It is hit or miss lately. Very very disappointing
3 Likes
What will happen if it decreases to 0? What would you call that? Could they decrease it any further?
Video requires the AI to anaylize every frame and vortex of the image and that is a LOT of compute. Those are the ones that should be rate limited and less for text and still images or videos posted that are already predone where the AI doesn’t have to ‘generate’ any moving images.
I’d like an update on the Google One integration when’s it going to happen we’ve been waiting for weeks and frankly I want to get back to using the Studio again.
7 Likes