Gemini API 503 The Model Is Overloaded

How to fix it

Try these in order — the most common fix is first.

1. 1. Retry with exponential backoff

   Copy steps
   1. Wait a few seconds, then retry with increasing delays and jitter
   2. Cap retries at a handful so you do not hit rate limits
   Did this fix it? Yes No

2. 2. Fall back to another model

   Copy steps
   1. On repeated 503s switch to a Flash or earlier stable model
   2. Keep a fallback model name in your config
   Did this fix it? Yes No

3. 3. Check status and report

   Copy steps
   1. Check the Gemini API / Google AI Studio status page
   2. Search discuss.ai.google.dev for same-day reports
   3. Use Send feedback in AI Studio if it persists for hours
   Did this fix it? Yes No

4. 4. Run less busy

   Copy steps
   1. Reduce prompt size and parallel requests
   2. Move batch jobs off peak hours
   Did this fix it? Yes No

Symptoms

Requests fail with 503 UNAVAILABLE and 'The model is overloaded', often on newer or popular models such as Gemini Pro previews, usually at peak hours.

Why it happens

• Google capacity for that model is saturated, especially new or preview models
• Peak-hour demand
• Paying does not exempt you; paid users report it too
• Aggressive retries can then turn into 429 errors

This guide also covers

• gemini 503 overloaded
• gemini model is overloaded
• gemini api 503 unavailable
• gemini overloaded please try again later

Other Google Gemini errors

• Gemini Something Went Wrong Error
• Gemini Image Generation Not Working
• Hey Google Not Working After Switching to Gemini
• Gemini Usage Limit Reached
• How to Cancel Google Gemini Subscription (Google AI Plus, Pro, Ultra)
• Gemini API Error 429 Resource Exhausted
• Gemini Says I Can't Help With That
• How to Turn Off Gemini on Android