Parameters
Streaming
Streaming sends tokens to the client as they are generated rather than waiting for the complete response.
It does not make generation faster. It changes what the user waits for — from the whole response to the first token — which is commonly a five to ten times improvement in perceived latency for no cost beyond handling the stream.
Time to first token is dominated by prefilling your input, so shortening the prompt improves felt speed as well as cost.
In practice
Streaming does not make anything faster, it changes what the user waits for. A 400-token answer at 55 tokens per second takes about 7.3 seconds either way; streaming turns that into a 0.6-second wait followed by text arriving faster than most people read. Time to first token is the number that matters, and long input drives it through prefill.
Common questions
Does streaming make responses faster?
No, it changes what the user waits for. Total time is the same; the wait before anything appears drops from the full generation to time to first token, which is usually a tenfold improvement in perceived speed.
What breaks when streaming?
Abandoned streams that are never aborted keep generating and keep billing. Always pass an abort signal and cancel it when the client disconnects, or you pay for answers nobody is reading.