Pricing
Pay for completed work, not idle infrastructure.
Common Compute is optimized for batch inference workloads.
Positioning
Pricing depends on workload type, dataset size, completion time target, and verification level.
Common Compute is built for batch inference workloads. The right price depends on the job, the amount of data, and how strict the verification path needs to be.
Example pricing table
Example benchmarks by workload type.
| Workload | Fit | Current benchmark |
|---|---|---|
| Embeddings | Large document and retrieval indexing runs | Request estimate |
| Transcription | Audio archives and scheduled media pipelines | Request estimate |
| OCR | Scanned PDFs and document extraction batches | Request estimate |
| Preprocessing | Dataset cleanup and pipeline preparation | Request estimate |
Get started