API

Streaming

Set stream: true to receive the response incrementally over SSE.

With "stream": true in the request body, the server responds with server-sent events. Each line starts with data: followed by a JSON chunk; the final line is data: [DONE].

Event stream

data: {"id":"chatcmpl-3f9a…","object":"chat.completion.chunk","model":"deepseek-v4.1-flash-heretic","choices":[{"index":0,"delta":{"role":"assistant","content":"Hel"}}]}

data: {"id":"chatcmpl-3f9a…","object":"chat.completion.chunk","model":"deepseek-v4.1-flash-heretic","choices":[{"index":0,"delta":{"content":"lo"}}]}

data: {"id":"chatcmpl-3f9a…","object":"chat.completion.chunk","model":"deepseek-v4.1-flash-heretic","choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":{"prompt_tokens":9,"completion_tokens":2}}

data: [DONE]

All chunks of one request share the same id. Concatenate choices[0].delta.content from each chunk to get the full reply.

Python

stream.py

stream = client.chat.completions.create(
    model="deepseek-v4.1-flash-heretic",
    messages=[{"role": "user", "content": "Tell me a short story"}],
    stream=True,
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

Notes

Lines that start with a colon are keep-alive comments. Ignore them.
If an error occurs mid-stream, you receive a data line containing an error object in the same format as non-streaming errors.
For reasoning models, the reasoning may arrive in delta.reasoning_content while the answer stays in delta.content.