Runpod Serverless Flex — B200 180GB — Event History
Provider: RunPod
Same native USD/GPU-hour on-demand Serverless Flex context: 8.64 to 10.79 (+24.88%) on the current Runpod public rate card, observed 2026-10-29. Effective date unknown; rate-card header remains October 8. Not a Batch Inference discount or FX change. Account-specific invoice not checked.
Runpod Batch Inference GA announced 2026-10-28: existing queue-based Serverless endpoints can submit asynchronous batches on dedicated batch workers. For B200, Runpod advertises 20% off on-demand GPU compute time for batch requests, with non-guaranteed starts, up to 5,000 requests/batch, 10 active batches/endpoint and default 50,000 requests/day/endpoint. Load-balancing endpoints excluded. This is a new batch capability/terms event, NOT a PRICE DROP: on-demand B200 180GB remains USD 8.64/GPU-hour. Exact invoice-level batch rate not verified; technical docs direct users to sales for rates. https://docs.runpod.io/serverless/batch-jobs
NEW canonical Serverless resource. Runpod lists NVIDIA B200 180GB Serverless Flex capacity at $8.64/GPU-hour with per-second billing.