Facing issue: Gemini Batches stuck in "JOB_STATE_PENDING" for 20+ hours

Gemini API
, ,
1

Hi Google team,

I am encountering an issue with the Gemini Batch API in which a submitted job has remained in the JOB_STATE_PENDING status for more than 20 hours.

Usually, batch executions complete within approximately two to three hours. This prolonged delay is therefore unexpected and concerning, particularly as it is approaching the standard 24-hour processing threshold.

The job was executed using the gemini-2.5-flash model. I have tested both smaller workloads and higher-volume requests; however, all submissions continue to remain in a pending state without progress.

Could you please help clarify the potential cause of this behavior and advise on any known service delays or constraints that may be impacting batch processing?

I would appreciate your guidance and any updates you can provide.

Thank you for your time and assistance.

1 Like
3

Same issue happened on my side as well since Jan 7 midnight, and it is still unsolved now. I have to quit batch api.

Correction: it works now.

4

Hi @Nakul_M1,

Thanks for your report. We’ve pushed a fix that should resolve the problem. Please check and let us know if the problem persists.

5

Hi @yyyyrz,

Thanks for the confirmation.

7

Hi @chunduriv , Yes it is working for me now. Thank you.

9

Hi guys,

Whatever fix you applied. I am having the same issue. Should I cancel all the jobs?

10

In general, the main reason jobs get stuck in pending is that the servers are under high load. The Batch API is treated as best effort, so if there is high load from non batch traffic, it will keep waiting for traffic to dip. If that never happens, the batch will time out. For 2.5 Flash-Lite, we rolled out a change which will hopefully help some batches go through faster.

11

Hi Logan, all my batches are now expired and not completed. Do you know if we are charged for the expired batches? Because if we are being charged for the expiring batches, then there’s no reason the use the batch at all. It is frustrating.

12

@chunduriv, the issue still seems to persist and its been 72 hours and the batch jobs are still in JOB_STATE_PENDING, can you please check and provide any update for the same
Model: gemini 3 flash preview

13

the issues seems not been resolved to me. I submitted a task and it remained in JOB_STATE_PENDING status more than 20 hours

14

same here, right after I switched to gemini 3.1-pr. I am switching back to 3-pro until it gets deprecated, hoping the issue will be solved by then

15

Have also been stuck on JOB_STATE_PENDING for over 24 hours with Gemini-3-flash-preview in batch mode. Had previously been running within 20 minutes. Issue seemed to start last week.

16

Same issue here. Was working fine up until about a week ago - now, even if I make the requests super small, just sits in pending forever.

17

same issue … Gemini-3-flash-preview is unusable for batch api right now … needed to switch to gemini-2.5-flash … when is it expected to be fixed?

18

Same issue here as well. My current job has been running for about 55 hours now, and still stuck in pending status. Might as well cancel it and rerun

My model: gemini-3-flash-preview

19

lol is someone going to look into this ? this is a MAJOR issue and seems like noone cares.. I cant process a single batch using gemini-3-flash-preview model..

2 Likes
20

Hi @chunduriv ,I had reported this issue previously, and I am now encountering it again. The Gemini Batch API job has been in the JOB_STATE_PENDING status for over 45 hours, which significantly exceeds the usual processing time.

Could you please look into this and share any updates?

Thank you for your time.

21

I face same problems, batches with gemini-3-flash-preview just never deliver results during 24 hours. It is super slow for those that are lucky to response with something different than JOB_STATE_PENDING. Is there anything can be done about it?