The new metrics provide visibility into key performance indicators: TimeToFirstToken measures latency from request to first token output in streaming APIs, assisting SLA monitoring without client changes, while EstimatedTPMQuotaUsage tracks token usage including caching effects across all inference APIs. Both update every minute for successful requests.
Available across commercial Bedrock regions, these metrics help users detect latency degradation early and proactively manage quota consumption to avoid rate limiting. Integration is seamless, requiring no API changes or opt-in, streamlining operational monitoring for generative AI workloads.
Source & further reading
Amazon Bedrock now supports observability of First Token Latency and Quota Consumption ↗Source published 2026-03-10. This report was prepared with AI assistance from the linked primary-source material. Claims and limitations should be checked against the original announcement or research.