Gemini too many requests: why videos trip it, and 6 fixes

How do you fix Gemini's too many requests error?

Run these six steps in order. None of them deletes a chat, and the first three take a minute.

1. Stop resending for a few minutes. The message only says "Try again later", and every retry is one more request. Copy your prompt out of the box first.
2. Switch the model from Pro to Flash and send the same prompt again. In a March 2026 thread, a Pro prompt failed for three days while the Fast and Thinking models "gives the answer immediately".
3. Pause video and image generation. In a July 2025 thread, a reply says normal chats and Deep Research still worked while video kept failing.
4. Open Settings, then Usage Limits, at the bottom left of gemini.google.com. A refresh time there means you hit the usage limit, and Gemini usage limit reached covers it.
5. Check the Gemini row on Google's Workspace Status Dashboard. It's a Workspace page, but its 10 June 2026 Gemini incident hit personal accounts too.
6. Send feedback from Settings and help if the lock survives a day. Include a screenshot, the exact message and when it started.

Gemini shows the message in two wordings. Screenshots posted to Google's Gemini Apps Community in 2026 read "Too many requests in a short time period. Try again later." (iPhone, April) and "Too many requests in a short time period. Please try again later." (Android in March, iPhone in July). It appears as a small banner over the chat box. It names no model and no time.

We searched the Gemini Apps Community) on 2 November 2026 and kept every thread whose title or first post quotes "too many requests" in the Gemini app. API threads were left out. That left 19, opened from 2 June 2025 to 4 July 2026, and 12 of them came in June and July 2025. Where a post didn't say what the poster was doing, we read the screenshot. Two showed a one-word text prompt ("Hi" and "Ftre"), and one showed the Video button switched on.

Which too many requests message are you seeing in Gemini?

The wording tells you which wall you hit. Google publishes some of these strings and not others. This table covers the Gemini app only, with where each one was reported. API errors have their own table further down.

The last two rows are the tell. Google's Gemini Apps limits page says that when you reach your limit, "Gemini Apps gives you another notification that tells you when your limit will refresh." A video limit message from May 2025 did exactly that. "Too many requests in a short time period" never does, and the limits page doesn't mention it.

If the banner reads "Something went wrong" with a number instead, that's a different error, covered in Gemini something went wrong.

Why does Gemini say too many requests when you barely used it?

Gemini says it because the lock doesn't count your prompts the way the message suggests. In the 19 threads, a light user got it on a follow-up question after six idle days. Two more got it on a one-word prompt. Google hasn't published what this lock measures, so the threads are the only record of what trips it.

Video generation. Seven posters were making a video, more than any other group. Google's limits page explains why video is heavy: "Video generally requires the most usage." In June 2025 a poster wrote that the message kept coming back "for over 24 hours, even though I've stopped using the tool and waited." That thread has 87 upvotes, the most of the 19.

The Pro model. Three posters hit it with text prompts on Pro. One saw it on two Gmail accounts on the same phone from one morning. Another said Fast and Thinking answered the same prompt at once. So the lock can sit on one model and leave the others open.

Google outages. Only one of the 19 lines up with a known incident. On 10 June 2026 a poster got error 1076 on the web and error 13 on the phone, then "Too many requests" on every retry. That's the day Google's status dashboard logged a 50% error rate on Gemini prompts. So retrying through an outage can stack one error on top of another.

Clearing data rarely helps. Posters tried cache clearing, reinstalls, incognito windows, other devices and other accounts, and none of the 19 threads reports that one of those fixed it. Google's limits page does warn that "Limits may change without notice, including due to capacity constraints." That's the closest Google comes to explaining a lock on a quiet account.

How long does Gemini's too many requests error last?

Gemini's lock has no published length. The message only says "Try again later", and Google prints no time for it. Volunteers answering in the threads guess anything from a few minutes to an hour, but 6 of the 19 posters said it lasted 8 hours or more. Here's what posters reported in their own threads.

So give it a few minutes, then change what you were doing. Move from Pro to Flash, or from video to text. A lock that follows you to another model and survives a day isn't yours to fix. Report it with Send feedback in the Settings and help menu, which Google's feedback page describes.

Does paying for Google AI Pro stop too many requests?

No. Four of the 19 posters said they paid for a Google AI plan, and a fifth was using the few Pro prompts a free account gets. One paid poster, on Google AI Pro, made about three videos and was locked out for two days. Another, in a June 2026 thread, said ten video attempts failed with "too many requests" while the usage bar still went down. That was a week after Google said on 28 May 2026, per 9to5Google: "If a request fails, you won't be charged." If your bar drops on failed videos, screenshot it and send it with your feedback.

Whizi, which publishes this page, runs the same kind of guard and prints the numbers: 10 chat messages a minute and 60 an hour per user, the same on every plan. Its error reads "Too many requests. Please wait and try again." and a refused request spends none of your allowance (Whizi rate limits). Gemini sits in the same picker as Claude and GPT (Gemini in Whizi). Where Whizi loses: 60 messages an hour is a hard ceiling, and its import takes ChatGPT and Claude exports but not your Gemini history.

How do you fix 429 RESOURCE_EXHAUSTED on the Gemini API?

Wait and retry with exponential backoff, then find out which limit you hit. A Gemini API 429 means your project passed a requests, tokens or spend limit, or Google ran short of capacity for that model. Google's rate limits page (updated 28 October 2026) says limits are measured in requests per minute, tokens per minute and requests per day. It adds: "Rate limits are applied per project, not per API key." A new key in the same project won't help. Your project's live limits and tier are on AI Studio's rate limit page, and the same Google page links a form to request a paid tier increase.

Three details catch people out. Requests per day "reset at midnight Pacific time", so a daily cap won't clear in an hour. "Rate limits are more restricted for experimental and preview models", which fits the Tier 3 threads above. And spend limits run on a rolling 10 minute window: $10 on Tier 1, $50 on Tier 2 and $200 on Tier 3. Tier 2 needs $100 spent plus 3 days, and Tier 3 needs $1,000 plus 30 days.

You may not need to write the retry loop. Google's troubleshooting page says the Python SDK already retries up to four times, from about 1 second up to 60 seconds. Google's API errors page lists the too_many_requests code with almost the app's words. That code belongs to the Interactions API. On Vertex, Google's 429 page adds one more fix: use the global endpoint instead of a regional one.

Gemini CLI shows a bare "429 Too Many Requests" as well. In an April 2026 GitHub issue, a Google AI Pro subscriber signed in with Google got it on version 0.36.0. A Gemini CLI maintainer closed it as a duplicate three days later. The CLI's daily request quotas, and what happens when Pro runs out, are in Gemini usage limit reached.