A point-of-sale screen must answer within hundreds of milliseconds.
Compare quality, price, and speed.
Choose two things to measure or consider.
Measure p95 latency and timeouts on representative inputs, and consider a smaller model or shorter output when it meets quality requirements.
Detailed explanation
Percentiles reveal slow requests that averages hide.
Reducing unnecessary model work can improve speed and cost.
Tail latency and errors matter for user experience.
Model size does not guarantee speed.
Check official documentation
Compare the following points with the official documentation.
Amazon Bedrock公式モデル比較・料金とAIF-C01 Domain 3.1の選択観点を確認する。
Official references
What to confirm
平均だけでなく分位点と品質・費用を合わせてモデルを選べる。
Key points
- p95
- Quality
- Cost
Notes
- Environment: AWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
- Compare the definition with the AIF-C01 objective and the service documentation before applying it.
- Keep the concept separate from a particular answer choice or implementation step.
Foundation review
p95 latency
The response time at or below which 95 percent of measured requests complete.
Timeout rate
The proportion of requests that exceed a client or service time limit.