Streaming

Standard server-sent events: set stream: true. Ask for stream_options.include_usage and the final chunk carries the same token counts a non-streaming response would, which is also exactly what you are billed.

stream = client.chat.completions.create(
    model="claude-sonnet-5",
    messages=[{"role": "user", "content": "Count to ten"}],
    stream=True,
    stream_options={"include_usage": True},   # final chunk carries token usage
)

for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="")
    if chunk.usage:
        print("\n", chunk.usage)

What passes through untouched

We proxy, we do not reinterpret. All of this reaches the model exactly as you sent it:

• Tool calling and structured output
• Vision inputs and system prompts
• Prompt caching
• Sampling parameters and stop sequences