Is the Gemini API down?
Google Cloud's live Vertex AI Gemini incidents are below; the Gemini API with an AI Studio key reports at aistudio.google.com/status. During an outage, requests fail with HTTP 503 "UNAVAILABLE" or HTTP 500. Retry with backoff, switch model, or use the other of AI Studio and Vertex AI.
Live Gemini status
Checked
No open Vertex Gemini API incidents
- Gemini API with an AI Studio keyCheck AI Studio status
- Last incident: Vertex AI Gemini API customers experienced increased error rates when accessing the global endpoint.resolved Feb 27
Source: status.cloud.google.com, Google Cloud's incident feed for Vertex AI. Read on our server about every 2 minutes.
Which Google status page covers what
| You call | Where to check |
|---|---|
| The Gemini API with an AI Studio key | aistudio.google.com/status. It has no public feed, so it is linked, not read live |
| Gemini on Vertex AI, in a Google Cloud project | status.cloud.google.com, read live above |
Which error is Google, and which is you
| What you see | What it means |
|---|---|
HTTP 503, UNAVAILABLE | The service is temporarily overloaded or down. Retry with exponential backoff |
HTTP 500, INTERNAL | An unexpected error at Google. Retry; if it repeats, check the status pages |
HTTP 504, DEADLINE_EXCEEDED | The request did not finish in time. Raise or remove the client deadline |
HTTP 429, RESOURCE_EXHAUSTED | Your quota or rate limit. See the Gemini quota fix |
Five minutes into a Gemini outage
- Check both status pages: AI Studio and Vertex AI report incidents separately.
- Retry with exponential backoff. Google's official SDKs include automatic retry logic.
- If one model is overloaded, switch to another Gemini model or to another vendor's model with the steps below.
- If you call one of AI Studio or Vertex AI, the other is a second route to the same Gemini models.
How to switch model in your agent
| Agent | Steps |
|---|---|
| Gemini CLI | Type /model and pick another Gemini model |
| Cline, Roo Code, Kilo Code | Open the extension's settings, change API Provider or Model, and save. The next request uses it |
| OpenCode | Type /models and pick a model from another provider, or start with opencode -m provider/model |
| Aider | Type /model followed by another model name; the chat history stays |
| Your own code | Catch 5xx and 529, retry with exponential backoff, then send the request to a fallback model |
What Standard Compute does when Gemini fails
- Every request goes through OpenRouter. Gemini models are served by Google AI Studio and Google Vertex AI; if the preferred one fails, OpenRouter can send the request to the other for the same model.
- Each failed call is retried twice on a 429, a 5xx or a network error, after 0.5 and then 1 second, or after the provider's Retry-After up to 8 seconds. If every attempt fails, your agent gets HTTP 503 with "The model is temporarily unavailable. Please retry shortly."
- A specific model ID always stays that model; failover only changes the provider. Only the auto-router, the model
standardcomputewith no model selection, can switch to a different model.
The catch
Both routes are Google. If a Gemini model fails on AI Studio and Vertex AI together, Standard Compute cannot serve it either; switch to another model on the same key, such as anthropic/claude-sonnet-5.5 or openai/gpt-6-astra, or wait.
Where Standard Compute fits
Standard Compute is an OpenAI- and Anthropic-compatible LLM gateway for coding agents at https://api.stdcmpt.com/v1, with monthly plans from $19 (USD) and a fixed compute budget. One key serves Claude, GPT, Gemini, DeepSeek and GLM models, so moving off a struggling model during an outage takes one changed model ID.
FAQ
Where do I check if the Gemini API is down?
For the Gemini API with an AI Studio key, check aistudio.google.com/status. For Gemini on Vertex AI, check status.cloud.google.com, where incidents list the Vertex Gemini API as the affected product.
What does 503 UNAVAILABLE 'The model is overloaded' mean?
Google does not have capacity for that model right now; your quota is fine. Retry with exponential backoff, and switch model if it lasts.
Is a Gemini 429 RESOURCE_EXHAUSTED an outage?
No. It means your project hit a rate limit or quota. Wait for the window to reset, lower concurrency, or raise the quota.
Can a gateway keep Gemini requests working during an outage?
Only while one of Google's routes still serves the model. On Standard Compute a Gemini request can move between Google AI Studio and Vertex AI for the same model, with two retries, and returns HTTP 503 if every attempt fails.