Continuous 429 / RESOURCE_EXHAUSTED errors on Gemini API while on Tier 3, with no visible limits reached

[Lito]

Hello,

We are continuously running into limit-related errors on the Gemini API, even though we are currently on Tier 3.

So far today, we have already had more than 12,000 failed requests. According to the dashboard, it does not look like we have reached any of our limits, or at least we cannot see any clear indication of a specific quota being exceeded.

At the moment, we are seeing an overall error rate of around 75% across requests.

The errors we are receiving are mainly 429 responses with the status RESOURCE_EXHAUSTED. The full response is:

"error": {
    "code": 429,
    "message": "You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: ai.google.dev/gemini-api/docs/rate-limits. To monitor your current usage, head to: ai.dev/rate-limit. ",
    "status": "RESOURCE_EXHAUSTED",
    "details": [
        {
            "@type": "type.googleapis.com/google.rpc.Help",
            "links": [
                {
                    "description": "Learn more about Gemini API quotas",
                    "url": "ai.google.dev/gemini-api/docs/rate-limits"
                }
            ]
        }
    ]
}

What is confusing is that the error message points to a quota or rate limit issue, but the dashboard does not show any limit being reached, at least not in an obvious way.

We would appreciate help understanding:

1. How can we identify exactly which limit we are exceeding?

2. Could these errors be caused by limits that are not shown clearly in the dashboard?

3. Are there limits per model, region, minute, day, project, request type, or token usage that we may not be seeing correctly?

4. How can we check actual usage versus available limits for each model?

5. Is there a recommended way to monitor these errors and distinguish between exhausted quota, rate limiting, billing issues, or another restriction?

6. What steps should we follow to reduce this volume of errors or request a quota review/increase?

The current impact is significant: we are seeing thousands of failed requests and very high error rates, but we are not sure where to start diagnosing the issue.

Any guidance would be greatly appreciated.

Thank you.

[Jon_Matthews]

Hi Lito,

Please add your details to this form & we’ll check it out.

[jhero]

Hi Jon, we’re experiencing the same issues with 429 RESOURCE_EXHAUSTED and we’re on Tier 3 with over 50% error rate despite being nowhere close to the limits.

I filled out the form and haven’t heard back. We need a resolution for this as soon as possible – this is affecting our production environment and causing a lot of problems for our users.

[Lito]

We’ve identified the issue: the limit we were hitting was Gemini 3 Search grounding. The problem is that the usage statistics on the Gemini API Rate Limit page didn’t always show that the limit had been reached, which made it very difficult to spot the problem.

[Jon_Matthews]

Hi jhero, apologies for the delay. I’m actively working through these with engineering. To be clear: I’ll communicate back when I have answers, but I also want to get to the bottom of these reports to prevent these issues biting people in the future.

Since I can see from your form submission that you haven’t obviously breached any limits, here’s my advice for now:

• Most of the time, as in the post above, this is because of the limits on search grounding. That’s visible lower down in the dash.
• It’s possible to breach TPM limits without it showing up in the dash, just because of the granularity in the graph. Implementing backoff and retry should help with this.

Jon