AWS has introduced usage-based pricing for Amazon SageMaker AI bidirectional streaming inference. Through AWS Marketplace, model providers can now meter and charge customers according to actual consumption during a live stream, rather than through other billing arrangements. Deepgram is the launch partner, enabling usage-based billing for its speech-to-text and text-to-speech offerings on AWS Marketplace.

Why it matters

Usage-based metering for streaming inference aligns billing with consumption during live interactions, which is relevant for workloads such as real-time speech processing. This provides model providers on AWS Marketplace with a more granular way to charge for streaming services.

Who should care

Model providers distributing streaming inference models through AWS Marketplace and customers using bidirectional streaming workloads, including speech-to-text and text-to-speech applications via Deepgram, should take note of the new pricing option.