I have been encountering a persistent blocker all day today while trying to build an app.
The system constantly throws these two errors after running for a few seconds:
“Encountered retryable error from model provider: Agent execution terminated due to error.”
“An internal error occurred.”
Here is the exact context and the troubleshooting steps I have already taken to isolate the issue:
Model used: Gemini 3.7 Flash.
Initial state: I hit the “out of free generations today” limit earlier.
Action taken: I went to Google AI Studio, created a new API key, and successfully set up a Tier 1 (Prepay) billing account.
Account status: I added ~$40 to the Google Cloud account. The project dashboard clearly shows the “Pay per request” badge and the payment was confirmed.
Environment reset: I updated the API key in my platform’s environment secrets, hard-refreshed the workspace, and completely restarted the session.
Testing: I tried sending extremely simple prompts (to rule out complex context windows or infinite loops), but the agent execution is still terminated every single time.
Since my billing is active and the API key is freshly linked, is there currently a known outage with Gemini 3.7 Flash agents, or is there an additional backend GCP configuration I need to enable for third-party platforms to bypass this specific termination error?
Any insights would be highly appreciated. Thank you!
I am experiencing exactly the same issue in Google AI Studio Build with Gemini 3.7 Flash.
While free generations are available, Build works normally. As soon as the free generations are exhausted and AI Studio switches to “Pay per request”, Build becomes unreliable or stops working.
I then receive:
“Encountered retryable error from model provider: Agent execution terminated due to error.”
followed by:
“An internal error occurred.”
This also happens with very simple prompts, so it does not appear to be caused by the complexity of my project. The same project works normally again when free generations are available.
It takes AGES for Google AI Studio to even answer the simplist of questions. And bizarre: during de ‘free time’, it works like a charm. When free tims is up and I have to use my pay as you go, all the problems begin. Ik have a feeling to be paying for … less quality then the free version??
Could you please investigate whether there is an issue with the transition from free Build generations to the paid “Pay per request” execution path?
@Mahesh_Sutar I have reproduced this same issue specifically during the free-generations → Pay per request transition. I have Project ID, Billing ID, Applet ID, screenshots and failed task details available. Could you please DM me so I can provide them privately?