Skip to content

Qwen_3.8_flash

Model Overview

item value
Model ID (model) qwen3.8-flash
context window 1,000,000 tokens
Maximum Input 991,808 (Thinking Mode: 983,616)
Maximum Output 131,072
Max Chain-of-Thought 262,144
Input Modality Image, Text, Video
output modality Text
Default Thinking Mode Enable
Prefix Continuation Support
Context Caching Supported (explicit cache_control)

Request Protocol

https

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_Qwen_3.8_flash_20260826 (original model from the vendor: qwen3.8-flash)
messages - - - array[json] Yes Sequential chat messages
role - - string Yes system / user / assistant / tool
content - - string or array Yes Plain text is of type string; multimodal or explicitly cached content is of type array
type - string Yes (when it is an array) text / image_url / video / video_url
text - string Yes (type=text) Text content
cache_control - object No Explicit cache breakpoint,{"type":"ephemeral"}
image_url - object Yes (type=image_url) Image Information
url string Yes Image URL or Base64 Data URL
min_pixels - integer No Qwen3.8 Image Default/Minimum 65536
max_pixels - integer No Qwen3.8 image default 2621440, maximum 16777216
video - array Yes (type=video) A list of video frame URLs in the form of an image list
min_pixels - integer No Qwen3.8 has a default video frame setting of 65536, with a minimum value of 4096
max_pixels - integer No For Qwen3.8, the default video frame size is 655360, and the maximum is 2048000
total_pixels - integer No Qwen3.8 Video Total Pixels Default/Maximum: 819200000
video_url - object Yes (type=video_url) Video File Information
url string Yes Video URL or Base64 Data URL
min_pixels - integer No Same as the video type
max_pixels - integer No Same as the video type
total_pixels - integer No Same as the video type
partial - - boolean No Prefix continuation, default value is false; enabled for assistant messages
tool_calls - - array No Tool calls in assistant messages (multi-round postback)
id - string Yes Tool Response ID
type - string Yes fixed function
function - object Yes Tools and Parameters
name string Yes Tool Name
arguments string Yes JSON string input parameter
tool_call_id - - string Yes (tool role) The ID of the corresponding tool_calls
reasoning_content - - string No assistant historical thinking; when preserve_thinking is enabled, complete postback is required
stream - - - boolean No Whether to use streaming mode; defaults to false
stream_options - - - object No Streaming configuration, only takes effect when stream= true
include_usage - - boolean No The last chunk returns usage, which defaults to false
max_completion_tokens - - - integer No Recommended; total length of chain-of-thought + final answer; default/maximum value is the model's maximum output token count of 131072
max_tokens - - - integer No Deprecated soon; for new integrations, please use max_completion_tokens instead
temperature - - - float No Range [0,2); default thinking mode is 0.6, and it will be automatically adjusted to 0.6 if the value is lower than 0.6
top_p - - - float No Nucleus sampling, default temperature: 0.95; range (0,1.0]
response_format - - - object No Output format, default{"type":"text"}; Qwen 3.8 supports strict json_object
type - - string Yes text or json_object
tools - - - array No Function Calling Tool List
type - - string Yes Fixed function
function - - object Yes Tool Definition
name - string Yes Tool name, maximum 64 tokens
description - string Yes Tool Description
parameters - object No JSON Schema Parameter Description
tool_choice - - - string/object No auto/ none/ specify function; mandatory specification is not supported in thinking mode
parallel_tool_calls - - - boolean No Parallel tool invocation, false by default
enable_thinking - - - boolean No Mixed reasoning switch; enabled by default for qwen3.8-flash; can be placed at the top level of the body for direct HTTP calls
reasoning_effort - - - string No qwen3.8 default: xhigh; valid options: xhigh/ medium/ low;max/ high→ xhigh, minimal→ low, none→ disable thinking; other values will trigger an error; cannot be set together with thinking_budget
thinking_budget - - - integer No Maximum tokens for chain-of-thought; cannot be set simultaneously with reasoning_effort; supports mutual conversion and mapping with reasoning_effort
preserve_thinking - - - boolean No qwen3.8-flash defaults to true; the reasoning_content must be fully posted back, and must not be spliced into content
top_k - - - integer No Number of sampling candidate tokens, default is 20; non-OpenAI standard, use extra_body in SDK
repetition_penalty - - - float No Repetition penalty, default value for Qwen3.8 is 1.0
presence_penalty - - - float No Content repetition rate: default value in non-thinking mode is 1.5; valid range: [-2.0,2.0]
vl_high_resolution_images - - - boolean No High-resolution image, default false
tool_stream - - - boolean No Streaming output of complex tool parameters, only applicable when stream= true; qwen3.8-flash supports all modalities
enable_code_interpreter - - - boolean No The switch for the code interpreter on the Chat side, which defaults to false (different from the built-in code_interpreter tool in Responses)
stop - - - string/array No stop words
seed - - - integer No random seed, default 1234

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_Qwen_3.8_flash_20260826",
    "messages": [{"role": "user", "content": "Hello"}],
    "reasoning_effort": "none"
  }'

Response Example

{
  "id": "chatcmpl-xxx",
  "object": "chat.completion",
  "model": "MaaS_Qwen_3.8_flash_20260826",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "你好！",
      "reasoning_content": "..."
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 20,
    "total_tokens": 30,
    "completion_tokens_details": {"reasoning_tokens": 5}
  }
}

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_Qwen_3.8_flash_20260826",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true,
    "stream_options": {"include_usage": true}
  }'

SDK Invocation 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_Qwen_3.8_flash_20260826",
    messages=[{"role": "user", "content": "Hello"}],
    extra_body={"reasoning_effort": "none"}
)

OpenAI Responses

Request URL

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

Request Body Parameters

Parameter Name Secondary Parameters Three-level parameters Type Required Description
model - - string Yes Fixed as MaaS_Qwen_3.8_flash_20260826
input - - string or array Yes Plain text such as "Hello", or a message array
role - string Yes (when it is an array) user / assistant / system / developer
content - string or array Yes (when it is an array) Message content
type string Yes (content is an array) input_text / input_image etc.
text string Yes (input_text) Text content
image_url string Yes (input_image) Image URL or Base64 Data URI
stream - - boolean No Whether to use streaming mode; defaults to false
max_output_tokens - - integer No Qwen3.8: Combined upper limit for response + chain-of-thought; minimum 16
reasoning - - object No Thinking intensity; priority over enable_thinking
effort - string No Default: xhigh; 7 levels are supported in total: none/minimal/low/medium/high/xhigh/max;nonedisables reasoning;xhigh/ maxis only available in North China 2 and Singapore regions
enable_thinking - - boolean No Deprecated; it is recommended to use reasoning. effort; enabled by default for qwen3.8-flash
tools - - array No Built-in tools or custom functions; to use the code_interpreter, reasoning must be enabled (i. e., reasoning. effort is not set to none, or the default reasoning configuration is applied)
type - string Yes web_search / code_interpreter / function etc.
name - string Yes (function) Custom Tool Name
description - string Yes (function) Tool Description
parameters - object No (function) JSON Schema
tool_choice - - string/object No auto / none / required / allowed_tools object
previous_response_id - - string No ID of the previous response, valid for 7 days
instructions - - string No System instruction, start of inserted context
temperature - - float No Sampling temperature, range [0,2)
top_p - - float No Kernel sampling, range (0,1.0]
store - - boolean No Whether to store the response for subsequent reference; defaults to true

Request Example

curl -X POST 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer ${your_AK}' \
  -d '{
    "model": "MaaS_Qwen_3.8_flash_20260826",
    "input": "Introduce Qwen in one sentence",
    "reasoning": {"effort": "none"}
  }'

Response Example

{
  "id": "resp_xxx",
  "object": "response",
  "model": "MaaS_Qwen_3.8_flash_20260826",
  "status": "completed",
  "output": [
    {"type": "reasoning", "summary": [{"type": "summary_text", "text": "..."}]},
    {"type": "message", "role": "assistant", "content": [{"type": "output_text", "text": "..."}]}
  ],
  "usage": {
    "input_tokens": 12,
    "output_tokens": 48,
    "total_tokens": 60
  }
}

Streaming request example

curl -X POST 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer ${your_AK}' \
  -d '{
    "model": "MaaS_Qwen_3.8_flash_20260826",
    "input": "Hello",
    "stream": true,
    "reasoning": {"effort": "none"}
  }'

SDK Calling Method

from openai import OpenAI

client = OpenAI(
    api_key="${your_AK}",
    base_url="https://genaiapi.cloudsway.net/v1/ai/{endpointPath}"
)
response = client.responses.create(
    model="MaaS_Qwen_3.8_flash_20260826",
    input="Hello",
    reasoning={"effort": "none"}
)

Anthropic Messages

Request URL

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

Header (Additional to This Agreement)

Parameter Name Type Required Description
anthropic-version string Yes e. g. 2023-06-01

Request Body Parameters

Parameter Name Secondary Parameters Three-level parameters Type Required Description
model - - string Yes Fixed as MaaS_Qwen_3.8_flash_20260826
max_tokens - - integer Yes qwen3.8-flash: Combined cap for response + chain of thought
messages - - array[json] Yes Dialogue Message
role - string Yes user / assistant / system
content - string or array Yes Array of text or content blocks
type string Yes (array) text / image / video etc.
text string Yes (type=text) Text content
cache_control object No Explicit caching,{"type":"ephemeral"}
source object Yes (image/video) URL or Base64 source
system - - string or array No System prompt; the array format may include cache_control
type - string Yes (array) Fixed text
text - string Yes (array) System Prompt
cache_control - object No Explicit cache breakpoint
stream - - boolean No Streaming or not
thinking - - object No Deep thinking; enabled by default for qwen3.8-flash
type - string Yes enabled or disabled
budget_tokens - integer No Deprecated; use output_config. effort instead
output_config - - object No Vendor extension; SDK passthrough required
effort - string No qwen3.8 defaults to xhigh; available options are xhigh/ medium/ low;max/ high→ xhigh
format - object No Strictly structured output of json_schema
type string Yes Fixed json_schema
schema object Yes JSON Schema
tools - - array No Tool Definition
name - string Yes Tool Name
description - string No Tool Description
input_schema - object Yes Input JSON Schema
tool_choice - - object No auto / any / none / specify tool
temperature - - float No Range [0,2); default value of thinking mode is 0.6
top_p - - float No Nuclear Sampling
top_k - - integer No Size of Sampling Candidate Set
stop_sequences - - array No stop sequence

Request Example

curl -X POST 'https://genaiapi.cloudsway.net/{endpointPath}/v1/messages' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer ${your_AK}' \
  -H 'anthropic-version: 2023-06-01' \
  -d '{
    "model": "MaaS_Qwen_3.8_flash_20260826",
    "max_tokens": 4096,
    "thinking": {"type": "disabled"},
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Response Example

{
  "id": "msg_xxx",
  "type": "message",
  "role": "assistant",
  "model": "MaaS_Qwen_3.8_flash_20260826",
  "content": [{"type": "text", "text": "Hello!"}],
  "stop_reason": "end_turn",
  "usage": {"input_tokens": 10, "output_tokens": 15}
}

Example of streaming request

curl -X POST 'https://genaiapi.cloudsway.net/{endpointPath}/v1/messages' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer ${your_AK}' \
  -H 'anthropic-version: 2023-06-01' \
  -d '{
    "model": "MaaS_Qwen_3.8_flash_20260826",
    "max_tokens": 4096,
    "stream": true,
    "thinking": {"type": "disabled"},
    "messages": [{"role": "user", "content": "Hello"}]
  }'

SDK Invocation Method

import anthropic

client = anthropic.Anthropic(
    api_key="${your_AK}",
    base_url="https://genaiapi.cloudsway.net/{endpointPath}/v1"
)
message = client.messages.create(
    model="MaaS_Qwen_3.8_flash_20260826",
    max_tokens=4096,
    thinking={"type": "disabled"},
    messages=[{"role": "user", "content": "Hello"}]
)