Based on my tests it didnt happened with openrouter paid version but at the same time in both gemini cli and aider it is happening with free tier only and now its getting worse.
By the way, we cant blame google because of that! As a developer we didnt see any model near it in coding complex tasks and google let us use it for free for several months which was amazing for one of our projects. (Mabye around 1000 doller a month if we wanted to use paid version!)
We hope google fix it.
Edit
Because of lower server loads at around 7 pm utc it works better.
Thanks for all the efforts but I think these problems are all well known by Google. I think its not straight forward and easy to fix. My gut feeling tells me this is a problem with the model, not the API itself. Previous response by @Logan_Kilpatrick in this thread:
Logan is the Product Owner for the Google Gemini API and Google AI Studio.
@Rodger_Blom I can imagine that this issue is very non-trivial and no doubt hard, but they stated that this is a production ready model that they charge for, and that you can build your own project, product, or company with. They either need to move it back to a pre-release and stop charging, or fix it.
Hi,
Any update on this ticket?
I believe I got this problem inside my system but I am not pretty sure.
How do I reproduce this bug consistently ?
Is it by Sending multiple request at once from 1 IP address?
"[SYSTEM] Step [cognitive] on model [gemini-2.5-flash] returned an empty response (Attempt 18/50). Retrying in 3s..."
Warning
"[SYSTEM] Step [cognitive] on model [gemini-2.5-flash] returned an empty response (Attempt 19/50). Retrying in 3s..."
Warning
"[SYSTEM] Step [cognitive] on model [gemini-2.5-flash] returned an empty response (Attempt 20/50). Retrying in 3s..."
Warning
"[SYSTEM] Step [cognitive] on model [gemini-2.5-flash] returned an empty response (Attempt 21/50). Retrying in 3s..."
Warning
"[SYSTEM] Step [cognitive] on model [gemini-2.5-flash] returned an empty response (Attempt 22/50). Retrying in 3s..."
Warning
"[SYSTEM] Step [cognitive] on model [gemini-2.5-flash] returned an empty response (Attempt 23/50). Retrying in 3s..."
Error
"[SYSTEM] Step [cognitive] on model [gemini-2.5-flash] failed on attempt 24/50:" "{\"error\":{\"code\":429,\"message\":\"You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\",\"status\":\"RESOURCE_EXHAUSTED\"}}"
Error
"[SYSTEM] Step [cognitive] on model [gemini-2.5-flash] failed on attempt 25/50:" "{\"error\":{\"code\":429,\"message\":\"You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\",\"status\":\"RESOURCE_EXHAUSTED\"}}"
Thatâs not fair!
The quota is decreasing, but I am getting no responses! The rerunning prompt helped me somtimes after a few retries.
Hi @Bryan_Hughes thanks for starting this thread but I think we can conclude Google (@Logan_Kilpatrick) is not going to fix this. Rumors are building up on the release of Gemini 3. It is my believe that all energy is focused on the Gemini 3 release so they can burn this ship (Gemini 2.5) down.
Hi @Rodger_Blom, agreed, which is lame enough because while sure, if 3.0 is such a change that this issue goes away then why bother, but certainly doesnt give me any warm and fuzzy that this behavior will continue. I am pretty sure we are getting charged for the empty-responses but it is challenging to know for sure. I put in a feature request where the request should get back an identifier to final costs of the token usage relative to whatever rate you are at based on the token count. It would be nice to have this level of itemized billing - knowing that this request cost $.10 or $1.
But still 503 errors.