Quota exceeded for aiplatform.googleapis.com/generate_content_requests_per_minute_per_project_per_base_model — Imagen on Vertex AI API 429

Published 2026-09-23

HTTP 429 Retry depends on cause

The exact error

429 Quota exceeded for aiplatform.googleapis.com/generate_content_requests_per_minute_per_project_per_base_model with base model: imagen-3.0-generate. Please submit a quota increase request.

Your Google Cloud project hit a per-minute request quota that Vertex AI counts per base model. The error names both, which tells you which quota row to check.

What it means

The metric name reads left to right: generate-content requests, per minute, per project, per base model. In the Google Developer forums thread 181091, a user calling imagen-3.0-generate-002 from a GitHub workflow got this 429 with base model: imagen-3.0-generate. The error asks for a quota increase request.

This is a project quota error, not the shared-capacity 429 that Google’s error-code-429 guide lists for pay-as-you-go (“Resource exhausted, please try again later.”). It names a quota your traffic exceeded, or that Vertex believes it exceeded. The forum thread is mostly about the second case. The reporter says the project quota showed a value of 10 with none used, and billing was enabled. Two later posters report the same pattern: image generation worked in the console UI while the API returned 429, and one says raising the quota did not help. No root cause was posted. A community reply suggested quota propagation delay, a wrong region or project, or a different service account in CI. Treat all of that as unconfirmed.

One detail from the thread is worth checking in your own code: the reporter called Imagen through GenerativeModel(...).generate_content, and the metric in the error is a generate-content metric. Whether that is why the quota did not match is not established in the thread.

Where Imagen stands now. You are most likely to meet this error in older code or while migrating. Google’s Imagen page for the Gemini API says Imagen models are shut down and no longer available there. On Vertex AI (now branded Gemini Enterprise Agent Platform), Google’s Imagen docs describe the Imagen endpoints, including imagen-3.0-generate-002, as deprecated, list them under “Discontinued endpoints” with gemini-2.5-flash-image as the replacement, and recommend updating before June 30, 2026 to avoid service disruption. The docs do not say whether those Vertex endpoints still answer today, so check in your own project rather than assuming.

A related report: “user has exceeded quota”. In Google AI Developers Forum thread 169602, a user reports “Failed to generate image: user has exceeded quota. Please try again later.” from Imagen 4 Ultra in Google AI Studio, after linking a paid API key, for two days running. That is the AI Studio surface, not a Vertex project, and the thread has no answer. It shares the symptom (paid access, quota error anyway) but not the metric. Since Google now says Imagen is shut down on the Gemini API, that path is one to migrate away from rather than debug.

How to confirm it’s this

HTTP 429, and the message names generate_content_requests_per_minute_per_project_per_base_model plus a base model: value.
Google’s Vertex quotas page describes checking generative media quotas on the Quotas and System Limits page by filtering on the metric name and on the model name as a dimension. It also warns the quotas do not display unless the Agent Platform API is enabled for the project.
Confirm the project and region your code uses match that row, especially in CI.

Fix

Find the exact quota row for this metric and base model, in the same project and region as the failing call.
Request an increase from that row if your real traffic needs it. The error and Google’s 429 guide both point to a quota increase request.
Smooth your traffic. Google’s 429 guide warns about acceleration limits and recommends ramping up gradually. Batch jobs that fan out many image requests at once are the most likely to trip a per-minute quota.
Check how you call the model. Make sure you use the documented image generation path for your model, not a text-generation client. The code in thread 181091 used the Generative AI module of the Vertex AI SDK, which Google’s Vertex deprecations page says was deprecated on June 24, 2025 and scheduled for removal on June 24, 2026, in favor of the Google Gen AI SDK.
Plan the migration. New work should target a current Gemini image model rather than a deprecated Imagen endpoint. Gemini image models have their own quota rows, so check those after you switch.

Retry or not

It depends. A per-minute quota refills, so an occasional 429 during a burst is worth retrying. Google’s 429 guide recommends truncated exponential backoff on pay-as-you-go. Keep the identical payload and cap the number of attempts.

If the very first request fails while the console shows the quota unused, as in thread 181091, backoff is unlikely to help: the posters there kept getting 429s. Stop after a few attempts, fail the task clearly, and check project, region, quota row and call path.

When it’s not this

“Resource exhausted, please try again later.” Shared capacity, not your project quota. See the resource-exhausted page.
generate_content_free_tier_requests, limit: 0. A Gemini API key with no free image allowance. Retrying never helps.
Veo quotas. Google’s quotas page lists a different metric for Veo.

Error text last verified against the sources below on 2026-09-23.