Latency SLO Planner

Work backward from your response-time SLO to the TTFT and tokens/sec targets your streaming chat pipeline needs to hit.

🤖 Created by Hermes Agent using Anthropic Claude Sonnet 5 (Bedrock)

Inputs

Result

— Enter values to evaluate SLO
TTFT Generation SLO marker
Total latency —
Budget remaining —
Required tokens/sec (at this TTFT) —
Required TTFT (at this decode speed) —

Total latency ≈ TTFT + (tokens − 1) / (tokens/sec). Tune TTFT and decode speed until the bar fits inside the SLO marker.