Hi Team,
All of my Batch API jobs are currently stuck in JOB_STATE_PENDING.
Model: Gemini 2.5 Flash
Account Tier: Tier 3
Project Number: 884078709…
This issue is affecting all batch submissions. Could you please investigate whether there is a service-side issue
+1, we’re seeing the identical issue in the same window, adding data in case it helps the team triage:
All batch jobs sit in JOB_STATE_PENDING and never get scheduled. They’re accepted cleanly (HTTP 200, no error field); batchStats shows pendingRequestCount equal to requestCount, i.e. zero processed.
It’s batch-specific, not the models. Interactive generateContent on the same models returns normally in about 2 to 3 seconds throughout.
Not concurrency or quota. A single 1-request job submitted into an otherwise-empty queue also stays PENDING.
Reproduced across two separate projects and multiple models (gemini-2.5-pro-preview-tts, gemini-3.5-flash, gemini-3.1-flash-lite-preview, gemini-3.1-pro-preview), so it spans accounts and models. The original report here being Tier 3 means it spans tiers too.
Jobs submitted around 05:50 to 06:45 UTC on June 3 completed in 1 to 6 minutes; everything from roughly 16:00 UTC onward has stayed PENDING or expired. That points at batch scheduling/capacity rather than request validation.
Could the team confirm whether there’s a service-side batch scheduling issue? It would also help a lot to surface batch queue status somewhere, the public status pages stay green during these stalls because interactive serving is healthy.
Thanks.