Streaming responses

Set stream: true to receive Server-Sent Events. Responses emits typed events; Chat Completions emits chunks with choices[].delta. Their consumers are not interchangeable.

Responses (new code)
Chat Completions (compatibility)

stream = client.responses.create(
    model="aliyun/kimi-k3",
    input="Explain streaming.",
    stream=True,
)

for event in stream:
    if event.type == "response.output_text.delta":
        print(event.delta, end="")

Common text events include response.created, response.output_text.delta, response.completed, and error.

stream = client.chat.completions.create(
    model="aliyun/kimi-k3",
    messages=[{"role": "user", "content": "Explain streaming."}],
    stream=True,
)

for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

A Chat Completions choices[0].delta parser will not parse Responses events, and a Responses event-type parser will not parse Chat Completions chunks.

Handle disconnects, timeouts, terminal events, and partial tool arguments without infinite retries.

Last modified on August 12, 2026

Introduction

Core capabilities

Reliability