Messages API

Call the Messages API to get the model-generated response data. This endpoint is compatible with the Anthropic Messages API format — you can use the Anthropic SDK directly or the same JSON shape.

This page lists only the currently confirmed supported fields. Do not pass fields that are not listed here.

Endpoint

POST https://api.stepfun.ai/v1/messages

When using the Anthropic SDK, set base_url to https://api.stepfun.ai — the SDK will automatically append /v1/messages, you don’t need to include /v1 manually.

Request Parameters

• model string required
  Model name. Use the public model name your account has access to, e.g. step-3.7-flash or step-3.5-flash.
• messages object array required
  Conversation history; at least one message.

  Show child attributes

  • role string
    Role name. Common values: user, assistant.
  • content string or object array
    Plain text, or an array of Content Blocks.

    Show child attributes

    • Plain text message string
    • Content Block array object array

      Show child attributes

      • Text block object

        Show child attributes

        • type string
          Always text.
        • text string
          Text content.
      • Image block object

        Show child attributes

        • type string
          Always image.
        • source object
          Image source. Supports URL or Base64.

          Show child attributes

          • URL: { "type": "url", "url": "https://..." }
          • Base64: { "type": "base64", "media_type": "image/png", "data": "..." }
      • Tool use block object (initiated by the model)

        Show child attributes

        • type string
          Always tool_use.
        • id string
          Unique ID for this tool call.
        • name string
          Name of the tool being called.
        • input object
          Arguments passed to the tool.
      • Tool result block object (returned by the caller)

        Show child attributes

        • type string
          Always tool_result.
        • tool_use_id string
          ID of the matching tool call.
        • content string or object array
          Tool execution result.
        • is_error boolean
          Whether the tool execution errored.
• max_tokens int required
  Maximum number of tokens to generate; must be greater than 0.
• system string or array optional
  System prompt. Either a string or an array of text blocks.
• tools object array optional
  List of tool definitions.

  Show child attributes

  • name string
    Tool name.
  • description string
    Tool description — helps the model decide when to use it.
  • input_schema object
    JSON Schema describing the tool input.
• output_config object optional
  Structured output configuration.

  Show child attributes

  • effort string
    Controls how much reasoning the model performs. Models supporting three reasoning levels accept low, medium, high; step-3.5-flash-2603 accepts only low and high.
• stream boolean optional
  Whether to stream the response; default is non-streaming.
• temperature float optional
  Sampling temperature, between 0 and 2.
• top_p float optional
  Nucleus sampling parameter, greater than 0 and at most 1.
• top_k int optional
  top-k parameter, between 0 and 500.
• stop_sequences string array optional
  Stop sequences — generation stops when any of these appear in the output.

Response

Non-streaming response

Content-Type: application/json

{
    "id": "msg_xxx",
    "type": "message",
    "role": "assistant",
    "stop_reason": "end_turn",
    "usage": {
        "input_tokens": 20,
        "output_tokens": 12
    },
    "content": [
        {
            "type": "text",
            "text": "I'm an AI assistant."
        }
    ]
}

Response fields

• id string
  Unique message ID.
• type string
  Object type, always message.
• role string
  Role name, always assistant.
• content object array
  List of content blocks. Typically text; in tool-calling scenarios may include tool_use.
• stop_reason string
  Why generation stopped. One of end_turn, tool_use, max_tokens.
• usage object
  Token usage statistics — at least input_tokens and output_tokens.

Streaming response

Content-Type: text/event-stream

Streaming uses standard SSE format. Each event has an event: line and a data: line; data: is JSON.

Common event types: message_start, content_block_start, content_block_delta, content_block_stop, message_delta, message_stop, ping.

When streaming tool-call arguments, content_block_delta.delta.type may be input_json_delta.

event: message_start
data: {"type":"message_start","message":{"id":"msg_xxx","type":"message","role":"assistant","model":"step-3.5-flash"}}

event: content_block_start
data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Hello"}}

event: message_delta
data: {"type":"message_delta","stop_reason":"end_turn","usage":{"input_tokens":20,"output_tokens":12}}

event: message_stop
data: {"type":"message_stop"}

Examples

Basic chat

python

from anthropic import Anthropic

client = Anthropic(api_key="STEP_API_KEY", base_url="https://api.stepfun.ai")

message = client.messages.create(
    model="step-3.5-flash",
    max_tokens=1024,
    system="You are an AI chat assistant provided by StepFun. You are fluent in English, Chinese, and many other languages. You answer user questions quickly and accurately while protecting user data.",
    messages=[
        {
            "role": "user",
            "content": "Introduce yourself in one sentence."
        }
    ],
)

print(message)

js

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
    apiKey: "STEP_API_KEY",
    baseURL: "https://api.stepfun.ai"
});

async function main() {
    const message = await client.messages.create({
        model: "step-3.5-flash",
        max_tokens: 1024,
        system: "You are an AI chat assistant provided by StepFun. You are fluent in English, Chinese, and many other languages. You answer user questions quickly and accurately while protecting user data.",
        messages: [
            {
                role: "user",
                content: "Introduce yourself in one sentence."
            }
        ]
    });

    console.log(JSON.stringify(message));
}

main();

curl

curl https://api.stepfun.ai/v1/messages \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $STEP_API_KEY" \
  -d '{
    "model": "step-3.5-flash",
    "max_tokens": 1024,
    "system": "You are an AI chat assistant provided by StepFun. You are fluent in English, Chinese, and many other languages. You answer user questions quickly and accurately while protecting user data.",
    "messages": [
        {
            "role": "user",
            "content": "Introduce yourself in one sentence."
        }
    ]
}'

{
    "id": "msg_01XFDUDYJgAACzvnptvVoYEL",
    "type": "message",
    "role": "assistant",
    "stop_reason": "end_turn",
    "usage": {
        "input_tokens": 35,
        "output_tokens": 20
    },
    "content": [
        {
            "type": "text",
            "text": "I'm an AI chat assistant by StepFun, ready to answer your questions in English, Chinese, and other languages quickly and accurately."
        }
    ]
}

Streaming response

python

from anthropic import Anthropic

client = Anthropic(api_key="STEP_API_KEY", base_url="https://api.stepfun.ai")

with client.messages.stream(
    model="step-3.5-flash",
    max_tokens=1024,
    messages=[
        {
            "role": "user",
            "content": "Introduce yourself in one sentence."
        }
    ],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

print()

js

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
    apiKey: "STEP_API_KEY",
    baseURL: "https://api.stepfun.ai"
});

async function main() {
    const stream = client.messages.stream({
        model: "step-3.5-flash",
        max_tokens: 1024,
        messages: [
            {
                role: "user",
                content: "Introduce yourself in one sentence."
            }
        ]
    });

    for await (const event of stream) {
        if (
            event.type === "content_block_delta" &&
            event.delta.type === "text_delta"
        ) {
            process.stdout.write(event.delta.text);
        }
    }

    console.log();
}

main();

curl

curl https://api.stepfun.ai/v1/messages \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $STEP_API_KEY" \
  -d '{
    "model": "step-3.5-flash",
    "max_tokens": 1024,
    "stream": true,
    "messages": [
        {
            "role": "user",
            "content": "Introduce yourself in one sentence."
        }
    ]
}'

Using output_config.effort

python

from anthropic import Anthropic

client = Anthropic(api_key="STEP_API_KEY", base_url="https://api.stepfun.ai")

message = client.messages.create(
    model="step-3.7-flash",
    max_tokens=1024,
    messages=[
        {
            "role": "user",
            "content": "Explain reinforcement learning in three sentences."
        }
    ],
    extra_body={
        "output_config": {
            "effort": "medium"
        }
    }
)

print(message)

curl

curl https://api.stepfun.ai/v1/messages \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $STEP_API_KEY" \
  -d '{
    "model": "step-3.7-flash",
    "max_tokens": 1024,
    "messages": [
        {
            "role": "user",
            "content": "Explain reinforcement learning in three sentences."
        }
    ],
    "output_config": {
        "effort": "medium"
    }
}'