A 2s chat latency budget should account for:

Question

Accepted Answer

Retrieval, auth, model TTFT/generation, tool calls, and network — budget each stage — Latency budgets itemize retrieval + auth + model + tools + network — not just the LLM.

Answer

Only model inference, ignoring retrieval and tools

Answer

Only CSS animations

Answer

Only DNS of unrelated domains

A 2s chat latency budget should account for:

Short Explanation and Infographic

Full Explanation