Streaming

Set stream: true on a chat completions request to receive the response as server-sent events.

Event format

The response has the content type text/event-stream.
Each event’s data is a JSON object with object: "chat.completion.chunk". The new text is in choices[].delta.content.
Reasoning arrives in delta.reasoning and delta.reasoning_details. Tool call arguments arrive in delta.tool_calls.
The stream ends with the line data: [DONE].

Usage on the final chunk

boolean

true adds a final chunk with usage, sent before data: [DONE]. Without it, the stream has no usage data. Billing is the same either way.

Read a stream

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["FORTYTWO_API_KEY"],
    base_url="https://api.fortytwo.network/v1",
)

model_id = "qwen/qwen3.8-27b"

stream = client.chat.completions.create(
    model=model_id,
    messages=[{"role": "user", "content": "Describe the tide in three sentences."}],
    stream=True,
    stream_options={"include_usage": True},
)

usage = None

for chunk in stream:
    if chunk.usage is not None:
        usage = chunk.usage
    for choice in chunk.choices:
        if choice.delta.content:
            print(choice.delta.content, end="", flush=True)

print()
print(usage)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.FORTYTWO_API_KEY,
  baseURL: "https://api.fortytwo.network/v1",
});

const modelId = "qwen/qwen3.8-27b";

const stream = await client.chat.completions.create({
  model: modelId,
  messages: [{ role: "user", content: "Describe the tide in three sentences." }],
  stream: true,
  stream_options: { include_usage: true },
});

let usage;

for await (const chunk of stream) {
  if (chunk.usage) {
    usage = chunk.usage;
  }
  const text = chunk.choices[0]?.delta?.content;
  if (text) {
    process.stdout.write(text);
  }
}

console.log();
console.log(usage);

Timeouts

A stream can be silent for up to 4 minutes, for example while the model reads a long prompt. Set the client read timeout above 4 minutes. A shorter timeout can close a stream that is still working.
If nothing arrives for 4 minutes, the stream ends with an upstream_timeout error event.
A request can run for up to 60 minutes — see Tiers and limits. A stream that reaches the limit also ends with upstream_timeout.

Errors during a stream

If the request fails before the first event, the response is an ordinary error with an HTTP status and a JSON body — see Errors. If it fails after the stream started, the last event is an error, and the stream closes. A timeout has the code upstream_timeout and the message “The request exceeded its time limit.”; any other failure has upstream_error:

{
  "error": {
    "message": "The request could not be completed.",
    "type": "server_error",
    "code": "upstream_error"
  },
  "request_id": "req_…"
}

A stream that ends without data: [DONE] and without an error event lost its connection. A stopped stream cannot be resumed; a new request is charged separately — see What is billed. The x-request-id response header identifies the request for support.

Get started

Using the API

Running in production

Support

Resources

Event format

Usage on the final chunk

Read a stream

Timeouts

Errors during a stream

​Event format

​Usage on the final chunk

​Read a stream

​Timeouts

​Errors during a stream

Event format

Usage on the final chunk

Read a stream

Timeouts

Errors during a stream