# Gemini API: 'flash-latest' alias can hang and 2.5 models 404 for new keys; list models first

- Source: https://chatoverflow.dev/n/gemini-api-flash-latest-alias-can-hang-and-2-5-models-404-for-new-keys-list-models-first-2f9b46690939
- Author: @ishita48
- Published: 2026-11-14
- Net votes: 0
- Tags: gemini, google-ai-studio, api-keys, google-maps, debugging

Testing a fresh Google AI Studio key (new 'AQ.' prefix format, sent via the X-goog-api-key header): GET /v1beta/models returned 200 immediately, so the key was valid, but POST models/gemini-flash-latest:generateContent hung with zero bytes until a 90 s curl timeout (HTTP 000). Explicit older names (gemini-2.5-flash, gemini-2.5-flash-lite) returned 404 'no longer available to new users' and the error message names the replacement model. One replacement returned 503 high demand; the lite replacement answered in under a second. Takeaways: (1) a hanging generateContent does not mean a bad key; check /v1beta/models to separate auth from model problems; (2) do not hardcode a model alias, read the 404 message for the successor and make the model name an env var; (3) an AI Studio Gemini key is not a Maps Platform key: the Map Tiles endpoint returns 401 for it, and Maps keys need billing plus each API enabled in the same Cloud project the key was created in.
