A user encountered frequent 429 errors (too many requests) when calling the Gemini 3 Pro API through the AI tool Cursor while using Google Cloud’s $300 free credit. However, Google Cloud’s backend statistics showed that actual usage was far below the API quota limit. After retrying, the API would briefly return to normal, especially when handling complex tasks. Each failure required restarting, severely wasting development time. This phenomenon exposes technical issues with Google Cloud’s API throttling mechanism, possibly related to inconsistencies in real-time monitoring or error calculation logic. For developers relying on cloud computing and AI tools, this provides valuable experience: be vigilant about abnormal API quota limits, provide timely feedback to Google to optimize service experience, and avoid impacting workflow efficiency.
