I’m seeing the same error across all of my older AI Studio Build apps since around Oct 16, 2026:
“Failed to call the Gemini API: Request contains an invalid argument.”
(ApiError 400 INVALID_ARGUMENT, thrown from ai.models.generateContent)
What I found:
- It affects ALL of my older apps, and the same code worked before with no changes.
- Remixing one of them under a different Google account produced the same error.
- All affected apps call Gemini directly from the browser (client-side, via @google/genai).
- A newer app that calls Gemini from a server-side Express endpoint (server.ts) works fine on the same account and free-tier quota.
- Changing the model name, the contents structure (file + text part), and the thinking config did NOT fix the browser-side call.
- Moving the call into a server-side endpoint (and switching to gemini-flash-lite-latest at the same time) fixed it. I changed the model and the call location together, so I can’t completely rule out the model, but the other changes had no effect while the call stayed in the browser.
So this looks related to client-side Gemini calls in Build apps rather than to the project, the account, or the quota. Could the team look into whether something changed in how browser-side calls are handled in the Build environment?
Happy to share more details privately.