A concrete example
When Claude writes, you see the words show up one at a time: that's streaming. Without it, you'd wait 10 seconds in silence.
Why it matters
Streaming cuts perceived latency. On a voice agent, it's actually a must-have.
You'll see it in chat interfaces where the words appear one after another.
Don't mix it up with
Latency: Latency is the total delay you feel between your request and the answer.
TTFT: TTFT measures the time between sending your request and the first token showing up.
Common mistakes
- Not turning streaming on for user interfaces (broken UX).
- Turning streaming on for a batch endpoint (pointless).
- Forgetting that streaming makes parsing JSON output trickier.
Quick checklist
- First I check whether the word names a concept, a tool, a risk, or a metric.
- I tie it to a concrete case: When Claude writes, you see the words show up one at a time: that's streaming. Without it, you'd wait 10 seconds in silence.
- I keep the main trap in mind: Not turning streaming on for user interfaces (broken UX).
Quick questions
What is Streaming in AI?
Streaming shows you the answer as the model generates it.
Where will I run into Streaming?
You'll see it in chat interfaces where the words appear one after another.
Which word should I read next?
Start with Latency, TTFT, TPS.