Streaming inference
Also known as: token streaming, SSE
Emitting partial outputs as tokens are generated - improves perceived latency for chat UIs while complicating logging and safety filters.
Contact if you need a term added for a security or procurement review.