MaaS_Stepfun

MaaS_Stepfun_step_5_preview_20260920

Model Overview

Item	value
Model ID (model)	step-5-preview
context window	1M tokens
Maximum Input	1M tokens
Maximum Output	64k tokens
input modality	text, images and videos
output modality	Text
Image	URL or Base64. Supported formats: JPG/JPEG, PNG, WebP, static GIF. Up to 60 images per upload. Detail level: low, high
video	stepfile:// from URL, Base64 or Files API. Supported formats: MP4, QuickTime, Matroska. For URL-sourced MP4 videos, the individual file size shall be less than 128MB, and the recommended duration is less than 5 minutes
Inference Intensity	low, medium, high. The Chat Completions API uses the reasoning_effort parameter; the Responses API uses the reasoning.effort parameter; the Messages API uses the output_config.effort parameter. The source document does not specify a default level.
Prompt Cache	Supported. The source document does not provide separate request parameters. The usage.prompt_tokens_details.cached_tokensof Chat and usage.input_tokens_details.cached_tokensof Responses return the tokens that hit the cache.
competence	Streaming output, function tool calling, JSON Mode, JSON Schema. Search and code execution are provided by the caller via tools, and the model does not directly access external services

Request Protocol

https

Header

Parameter Name	Type	Required	Description
Content-Type	string	Yes	application/json
Authorization	string	Yes	Bearer ${your_AK}

OpenAI Chat Completions

Request URL

POST https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/chat/completions

Request Body Parameters

Parameter Name	Secondary Parameters	Three-level parameters	Level 4 Parameters	Type	Required	Description
model	-	-	-	string	Yes	Fixed as MaaS_Stepfun_step_5_preview_20260920 (Vendor's original model: step-5-preview)
messages	-	-	-	array[json]	Yes	User input or model-generated messages to date
	role	-	-	string	Yes	system, user, tool, assistant
	content	-	-	string, array, or null	Yes	systemis text. useris text or a mixed list of images, text and videos. toolis the text result of function execution. assistantis text and can be null
		type	-	string	No	Multipart elements of user. text, image_url, video_url
		text	-	string	No	type= texttext at this point
		image_url	url	string	No	http/https image address, or data: image/jpeg;base64,${base64_string}. Supported formats: JPG/JPEG, PNG, WebP, static GIF. Maximum 60 images per request
		image_url	detail	string	No	low, high. highis interpreted based on the original image resolution, and tokens vary with the image size;lowis scaled to a fixed size, which saves more tokens
		video_url	url	string	No	The dedicated model chapter supports URL, Base64, or stepfile://. Supported formats include MP4, QuickTime, and Matroska. For URL-based videos, a single MP4 file should be smaller than 128MB, and the recommended duration is less than 5 minutes.
	tool_call_id	-	-	string	Conditions	role= toolis required. The function execution ID returned by the assistant in the previous round
tools	-	-	-	array[json]	No	List of function tools. Only documented types are function
	type	-	-	string	Yes	Always be function
	function	name	-	string	Yes	Only English letters, numbers and _-are allowed, and the recommended length is no more than 64 characters
	function	description	-	string	No	Function function description for model selection
	function	parameters	type	string	No	Generally it is object
	function	parameters	properties	object	No	key refers to the parameter name. Each parameter includes type (string, number, integer, object, array, boolean) and description
max_tokens	-	-	-	integer	No	Maximum tokens to generate. Defaults to INF (no restriction, determined by the model). The maximum output for model-specific chapters is 64k tokens. The total number of input and generated tokens is subject to the 1M context window limit.
temperature	-	-	-	float	No	Sampling temperature, ranging from 0.0 to 2.0. The default value is 0.5
top_p	-	-	-	float	No	Nucleus sampling. Default value: 0.9
n	-	-	-	integer	No	The number of responses generated for each input. The default value is 1, there is no upper limit, and it is recommended not to exceed 5.
stream	-	-	-	boolean	No	Whether to enable SSE streaming response. The default value is false.
stop	-	-	-	string or array	No	Stop immediately upon hitting any of the stop strings. Empty by default
frequency_penalty	-	-	-	float	No	Sort in descending order of occurrence frequency in the generated text. The default value is 0, with a range from -2.0 to 2.0
response_format	-	-	-	object	No	Output format. Default value:{"type":"text"}
	type	-	-	string	Yes	text, json_object, json_schema. json_object enables JSON Mode; json_schema constrains the structure according to the schema
	json_schema	name	-	string	Conditions	type= json_schemais required. Schema name
	json_schema	strict	-	boolean	No	Whether to strictly follow the schema. The default value is false
	json_schema	schema	-	object	Conditions	Required when type=json_schema. JSON Schema
reasoning_format	-	-	-	string	No	Format of the reasoning field. By default, generalis used and returned via reasoning. When set to deepseek-style, reasoning_contentis also available, and the contents of the two fields are identical. Optional values include generaland deepseek-style
reasoning_effort	-	-	-	string	No	Inference intensity of this model: low, medium, high. A higher value indicates deeper reasoning, which may take more time. The source document does not specify the default value.

Request Example

curl -X POST 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/chat/completions' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer ${your_AK}' \
  -d '{
    "model": "MaaS_Stepfun_step_5_preview_20260920",
    "messages": [
      {"role": "user", "content": "Please introduce yourself in one sentence."}
    ],
    "reasoning_effort": "medium",
    "max_tokens": 1024
  }'

Response Example

{
  "id": "b7b56af0-52a6-483f-a589-948182676a1b",
  "object": "chat.completion",
  "created": 1709893411,
  "model": "MaaS_Stepfun_step_5_preview_20260920",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "I am an assistant focused on programming and professional knowledge work.",
        "reasoning": "The user asked for a one-sentence introduction.",
        "reasoning_content": "The user asked for a one-sentence introduction."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 20,
    "completion_tokens": 30,
    "total_tokens": 50,
    "prompt_tokens_details": {"cached_tokens": 0},
    "completion_tokens_details": {"reasoning_tokens": 8}
  }
}

Example of streaming request

curl -X POST 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/chat/completions' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer ${your_AK}' \
  -d '{
    "model": "MaaS_Stepfun_step_5_preview_20260920",
    "messages": [
      {"role": "user", "content": "Please introduce yourself in one sentence."}
    ],
    "reasoning_effort": "medium",
    "max_tokens": 1024,
    "stream": true
  }'

SDK Calling Method

from openai import OpenAI

client = OpenAI(
    api_key="${your_AK}",
    base_url="https://genaiapi.cloudsway.net/v1/ai/{endpointPath}",
)

completion = client.chat.completions.create(
    model="MaaS_Stepfun_step_5_preview_20260920",
    messages=[{"role": "user", "content": "Please introduce yourself in one sentence."}],
    max_tokens=1024,
    extra_body={"reasoning_effort": "medium"},
)
print(completion.choices[0].message.content)

reasoning_effortis a vendor extension field, which is transmitted transparently via the extra_bodyparameter in the OpenAI SDK.

OpenAI Responses

Request URL

POST https://genaiapi.cloudsway.net/v1/ai/{endpoint}/responses

Request Body Parameters

Parameter Name	Secondary Parameters	Three-level parameter	Type	Required	Description
model	-	-	string	Yes	Fixed as MaaS_Stepfun_step_5_preview_20260920 (vendor's original model: step-5-preview). The Responses parameter page currently lists support for step-5-preview
input	-	-	string or array	Yes	Plain text is equivalent to a role= usermessage; or a time-ordered array of messages and events
	role	-	string	Conditions	Text message roles: user, assistant, system
	content	-	string or array	Conditions	Plain text, or an array of content blocks
		type	string	No	Content blocks: input_text, input_image, input_video. Postback items: function_call, function_call_output
		text	string	No	type= input_texttext at the time
		image_url	string or object	No	Image address string, or {"url": "...", "detail": "high"}. Recommended base64 data URL. detailis lowor highby model chapter. URL must be accessible by server level public network
		video_url	string or object	No	A video address string, or{"url":". ..", "detail":"low"}. The dedicated model chapter additionally supports Base64 and stepfile://; supported formats include MP4, QuickTime, and Matroska; the MP4 accessed via URL must be smaller than 128MB, and the recommended duration is less than 5 minutes. The source document does not list all possible values for the video detailseparately
	id	-	string	Conditions	type= function_callis a unique ID generated by the server, which must be postedback as-is in multi-turn conversations
	call_id	-	string	Conditions	function_calland function_call_outputpairing ID
	name	-	string	Conditions	type= function_callthe tool name
	arguments	-	string	Conditions	type= function_callJSON string parameters
	output	-	string	Conditions	type= function_call_outputis the tool execution result, which is recommended to be a JSON string
instructions	-	-	string	No	Top-level system instructions
stream	-	-	boolean	No	Whether to enable SSE streaming output. The default value is false.
temperature	-	-	float	No	Sampling temperature, ranging from 0.0 to 2.0. No default value is provided in the source document.
top_p	-	-	float	No	Kernel sampling. The source document does not provide the default value and value range
max_output_tokens	-	-	integer	No	This is the maximum output tokens for this response, which restricts both the inference process and the final output. When the budget is insufficient, statuscan be incomplete, and outputmay only contain reasoningwithout the final message. The maximum output for the model's dedicated chapter is 64k tokens
reasoning	-	-	object	No	Inference Configuration
	effort	-	string	No	This model supports low, medium, and high. The source document does not specify a default value.
tools	-	-	array[json]	No	Currently only function is supported
	type	-	string	Yes	Always be function
	name	-	string	Yes	It is recommended to use only English letters, numerals and _-
	description	-	string	No	function function description
	parameters	-	object	No	JSON Schema for function parameters
	strict	-	boolean	No	When enabled, the model output parameters strictly match parameters
tool_choice	-	-	string or object	No	Currently only the string "auto" is supported.
text	-	-	object	No	Text output format
	format	type	string	No	text, json_object, json_schema
	format	name	string	Conditions	type= json_schemais required
	format	strict	boolean	No	type= json_schemais available. When enabled, the output strictly matches the schema
	format	schema	object	Conditions	type= json_schemais required

Request Example

curl -X POST 'https://genaiapi.cloudsway.net/v1/ai/{endpoint}/responses' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer ${your_AK}' \
  -d '{
    "model": "MaaS_Stepfun_step_5_preview_20260920",
    "input": "Please introduce yourself in one sentence.",
    "reasoning": {"effort": "medium"},
    "max_output_tokens": 1024
  }'

Response Example

{
  "id": "resp_xxxxxxxxxxxxxxxx",
  "object": "response",
  "created_at": 1772624997,
  "completed_at": 1772624998,
  "model": "MaaS_Stepfun_step_5_preview_20260920",
  "status": "completed",
  "error": null,
  "incomplete_details": null,
  "output": [
    {
      "type": "reasoning",
      "id": "rs_xxxxxxxxxxxxxxxx",
      "summary": [],
      "content": null,
      "encrypted_content": null,
      "status": null
    },
    {
      "type": "message",
      "id": "msg_xxxxxxxxxxxxxxxx",
      "status": "completed",
      "role": "assistant",
      "content": [
        {
          "type": "output_text",
          "text": "I am an assistant focused on programming and professional knowledge work.",
          "annotations": []
        }
      ]
    }
  ],
  "usage": {
    "input_tokens": 14,
    "input_tokens_details": {"cached_tokens": 0},
    "output_tokens": 52,
    "output_tokens_details": {"reasoning_tokens": 0, "tool_output_tokens": 0},
    "total_tokens": 66
  }
}

Streaming Request Example

curl -X POST 'https://genaiapi.cloudsway.net/v1/ai/{endpoint}/responses' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer ${your_AK}' \
  -d '{
    "model": "MaaS_Stepfun_step_5_preview_20260920",
    "input": "Please introduce yourself in one sentence.",
    "reasoning": {"effort": "medium"},
    "max_output_tokens": 1024,
    "stream": true
  }'

SDK Calling Method

from openai import OpenAI

client = OpenAI(
    api_key="${your_AK}",
    base_url="https://genaiapi.cloudsway.net/v1/ai/{endpoint}",
)

response = client.responses.create(
    model="MaaS_Stepfun_step_5_preview_20260920",
    input="Please introduce yourself in one sentence.",
    max_output_tokens=1024,
    reasoning={"effort": "medium"},
)
print(response.output_text)

Anthropic Messages

Request URL

POST https://genaiapi.cloudsway.net/{endpoint}/v1/messages

Request Body Parameters

Parameter Name	Secondary Parameters	Three-level parameter	Level 4 Parameters	Type	Required	Description
model	-	-	-	string	Yes	Fixed as MaaS_Stepfun_step_5_preview_20260920 (vendor's original model: step-5-preview). The Messages parameter page lists it as a verified example
max_tokens	-	-	-	integer	Yes	Maximum tokens to generate, must be greater than 0. The maximum output for the model-specific chapter is 64k tokens. The source document does not separately specify the upper limit for this parameter.
messages	-	-	-	array[json]	Yes	At least one chat message
	role	-	-	string	Yes	Commonly used: user, assistant
	content	-	-	string or array	Yes	Plain text, or an array of Content Blocks
		type	-	string	No	text, image, tool_use, tool_result
		text	-	string	No	type= texttext at this point
		source	type	string	No	Image source. urlor base64
		source	url	string	Conditions	source. type= urlthe image address when
		source	media_type	string	Conditions	source. type= base64specifies the media type, for example image/png
		source	data	string	Conditions	source. type= base64image data
		id	-	string	Conditions	type= tool_usethe tool call ID at this point
		name	-	string	Conditions	type= tool_usetool name at the time
		input	-	object	Conditions	type= tool_useParameters passed to the tool
		tool_use_id	-	string	Conditions	type= tool_resultthe corresponding tool call ID
		content	-	string or array	Conditions	type= tool_resultexecution result
		is_error	-	boolean	No	type= tool_resultwhether an error occurred during execution
system	-	-	-	string or array	No	System prompt. A string, or an array consisting of text blocks
tools	-	-	-	array[json]	No	Tool Definition
	name	-	-	string	Yes	Tool Name
	description	-	-	string	No	Tool Description
	input_schema	-	-	object	No	JSON Schema for tool input parameters
output_config	-	-	-	object	No	Output configuration. The Anthropic SDK passes through via extra_body
	effort	-	-	string	No	Thinking depth of this model: low, medium, high. The source document does not specify the default value.
stream	-	-	-	boolean	No	Whether to return results in streaming mode; non-streaming by default
temperature	-	-	-	float	No	Sampling temperature, ranging from 0 to 2. No default value is provided in the source document
top_p	-	-	-	float	No	Core sampling: greater than 0 and less than or equal to 1. No default value is provided in the source document.
top_k	-	-	-	integer	No	Range: 0 to 500. No default value is specified in the source document.
stop_sequences	-	-	-	array[string]	No	Stop generation when any stop sequence is encountered

Request Example

curl -X POST 'https://genaiapi.cloudsway.net/{endpoint}/v1/messages' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer ${your_AK}' \
  -d '{
    "model": "MaaS_Stepfun_step_5_preview_20260920",
    "max_tokens": 1024,
    "messages": [
      {"role": "user", "content": "Please introduce yourself in one sentence."}
    ],
    "output_config": {"effort": "medium"}
  }'

Response Example

{
  "id": "msg_xxx",
  "type": "message",
  "role": "assistant",
  "model": "MaaS_Stepfun_step_5_preview_20260920",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 20,
    "output_tokens": 12
  },
  "content": [
    {
      "type": "text",
      "text": "I am an assistant focused on programming and professional knowledge work."
    }
  ]
}

Streaming Request Example

curl -X POST 'https://genaiapi.cloudsway.net/{endpoint}/v1/messages' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer ${your_AK}' \
  -d '{
    "model": "MaaS_Stepfun_step_5_preview_20260920",
    "max_tokens": 1024,
    "stream": true,
    "messages": [
      {"role": "user", "content": "Please introduce yourself in one sentence."}
    ],
    "output_config": {"effort": "medium"}
  }'

SDK Calling Method

from anthropic import Anthropic

client = Anthropic(
    api_key="${your_AK}",
    base_url="https://genaiapi.cloudsway.net/{endpoint}/v1",
)

message = client.messages.create(
    model="MaaS_Stepfun_step_5_preview_20260920",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Please introduce yourself in one sentence."}],
    extra_body={"output_config": {"effort": "medium"}},
)
print(message.content)