Batch API is taking longer than 24h (gemini-3.1-pro)

Gemini API
,
1

Hi everyone,

I wanted to ask if anyone else has experienced issues with the Batch API. According to the documentation, tasks should be completed within 24 hours (and usually much sooner). However, my current job has been running for about 35 hours now and is still stuck in the JOB_STATE_RUNNING status.

Does this happen often?

I am using the latest model: gemini-3.1-pro-preview. Even though it is labeled as a “preview” model, the documentation says it “may be used for production” so I expected it to be working properly.

Does anyone have more information on why this might be happening or how to resolve it?

Thank you

7 Likes
2

Ok, now it expired. So it never finishes. Why someting expire without any error message or additional info?

2 Likes
3

Same issue here as well. My current job has been running for about 52 hours now, and still stuck in pending status.

My model: gemini-3-flash-preview

4

I am also experiencing this exact issue.

Previously, my batch operations utilizing gemini-3-pro-preview completed without any complications. However, my current jobs using gemini-3.1-pro-preview have remained stuck in the running state for well over 24 hours.

This is causing a critical disruption to my workflow. Following the shutdown of gemini-3-pro, the gemini-3-pro-preview model is no longer operational, meaning I cannot revert to it. Because gemini-3.1-pro-preview is failing to yield any results or complete its run cycle, I have been forced to rely on gemini-2.5-pro or gemini-3-flash as temporary, suboptimal workarounds to maintain my operations.

I would greatly appreciate any updates from the team or a prompt resolution to this API delay.

2 Likes
5

Here’s my case. I used to use “gemini-3-pro-preview” and batch api jobs could be done. While 3.0 stopped on March 9, that model name might point to 3.1 since then. And my jobs never get done now. Sad

1 Like
6

Same issue.

My model: gemini-pro-latest

1 Like
7

I’m facng the same issue here. I hope the admin looks in to this asap :thinking:

my model: gemini-3.1-pro-preview

1 Like
8

I got the same issue. I am using the Gemini-3 Flash Preview model.

9

I’m having the same issue, it’s like they get stuck on running and they never complete.

10

appears to be affecting deep research via interactions api also, presume linked

11

I am facing the same issue. We need this resolved ASAP.

12

Hi
Could you DM me with your project numbers where batch API are taking longer than 24 hours

2 Likes
13

(post deleted by author)

14

I got the same issue. I am using the Gemini-3 Flash Preview model.

3 Likes
15

I sent you a DM with the number. Thank you for this.

16

how do you send a dm?

17

Click on his name and there’s a “message” button

18

Hi Mustan, as a new forum member I cannot send you a DM, but I experience the same problems (batches never complete on v3.1 pro, no matter how small and regardless of the thinking settings, but always completed within minutes on v3 pro).

I would appreciate if somebody can look into this soon.

1 Like
19

I’m having the same exact issues. Even when reducing the batch size it still fails. I’m using the gemini 3 flash preview model. People on Reddit are reporting the same issues.

1 Like
20

Hi @Mustan_lokhand,
As a new forum member I also can’t DM you. But I’m having the same issue since 3-pro-preview ended and I switched to 3.1-pro-preview. I have tried different batch sizes but no success.
Thanks

1 Like