MaaS_GP_5.4

Basic Information

Project	Description
Base URL	https://genaiapi.cloudsway.net/v1/
Authentication Method	Bearer Token (API Key)
Response Format	JSON
Request Format	JSON

New Feature Introduction

Similar to the previous GP-5 model, GP-5.4 supports custom tools, parameters for controlling the level of detail and inference capabilities, and a list of allowed tools. GP-5.4 also introduces multiple features that make it easier to build powerful agent systems, handle larger amounts of information, and run more reliable automated workflows:

tool_search in the API: GP-5.4 improves tool search functionality in large tool ecosystems by lazy-loading tools. This makes tools searchable, loads only relevant definitions, reduces token usage, and improves the accuracy of tool selection in actual deployments.
1 million token context window: GP-5.4 supports a context window of up to 1 million tokens, enabling easier analysis of entire codebases, long document collections, or extended agent traces in a single request. Please read the "1 million token context window" section for more information.
Built-in computer usage capabilities: GP-5.4 is the first mainstream model with built-in computer usage capabilities, enabling agents to directly interact with software to complete, verify, and repair tasks within the build-run-verify-fix cycle.
Native Compression Support: GP-5.4 is the first mainstream model trained to support compression, capable of achieving longer agent trajectories while retaining key context.

Model Capability List

Capability / Model	MaaS_GP_5.4	MaaS_GP_5.4_pro	MaaS_GP_5.4_mini	MaaS_GP_5.4_nano
Input Support	Text, Image	Text, Image	Text, Image	Text, Image
/Chat Completions	✅	❌	✅	✅
/Responses	✅	✅	✅	✅
Tool search	✅	✅	✅Charged by token usage	❌
Web search	✅	✅	❌	❌
Computer Use Support	✅	❌	✅	❌
Most Suitable	General-purpose tasks, including complex reasoning, extensive world knowledge, and intelligent agent tasks with large codebases or multiple steps.	Tricky problems may take longer to solve and require more in-depth reasoning.	A large amount of coding, computer usage, and agent workflows that still require strong reasoning abilities.	Simple high-throughput tasks where speed and cost are of utmost importance.

Request Parameters

Parameter	Type	Required	Description
`model`	string	is	Model Name
`messages`	array	is	Conversation Message List
`temperature`	float	No	Sampling Temperature (0-2), default 1.0
`top_p`	float	No	Nucleus sampling parameter (0-1), default 1.0
`max_tokens`	int	No	Maximum number of generated tokens
`stream`	boolean	No	Whether to stream output, default false
`presence_penalty`	float	No	-2.0 to 2.0
`frequency_penalty`	float	No	-2.0 to 2.0
`seed`	int	No	Deterministic Generation Seed

The following parameters are only supported when using GPT-5.4 and the inference difficulty is set to none:

temperature
top_p
logprobs

New`phase`parameter

For GP-5.4 processes with long-running operations or many tools in the response API, please use the assistant message phase field to avoid premature termination and other undesirable behaviors.

phase is optional at the API level, but we strongly recommend using it. It is used for phase: "commentary" intermediate assistant updates (such as prompt messages before tool calls) and phase: "final_answer" the final answer. Do not add phase to user messages.

Request Example

/responses

/responses: Non-streaming synchronous request

Curl Request

curl --location --request POST 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses' \
--header 'Authorization: Bearer {YOUR_API_KEY}' \
--header 'Content-Type: application/json' \
--data-raw '{
    "model": "gpt-5.4-mini",
    "input": "Question A: Answer A."
  }'

Python Request

import requests

url = "https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses"

payload = {
    "model": "gpt-5.4-mini",
    "input": "question"
}

headers = {
    "Authorization": "Bearer {YOUR_API_KEY}",
    "Content-Type": "application/json"
}

try:
    response = requests.post(url, json=payload, headers=headers)
    response.raise_for_status() 
    print(response.status_code)
    print(response.json())
except requests.exceptions.RequestException as e:
    print(f"Request error: {e}")

Return Example

{
    "top_logprobs": 0,
    "metadata": {},
    "presence_penalty": 0.0,
    "reasoning": {
        "effort": "none"
    },
    "usage": {
        "input_tokens_details": {
            "cached_tokens": 0
        },
        "input_tokens": 12,
        "output_tokens_details": {
            "reasoning_tokens": 0
        },
        "total_tokens": 36,
        "output_tokens": 24
    },
    "created_at": 1775556238,
    "store": true,
    "tools": [],
    "content_filters": [
        {
            "content_filter_results": {
                "self_harm": {
                    "severity": "safe",
                    "filtered": false
                },
                "jailbreak": {
                    "filtered": false,
                    "detected": false
                },
                "hate": {
                    "severity": "safe",
                    "filtered": false
                },
                "sexual": {
                    "severity": "safe",
                    "filtered": false
                },
                "violence": {
                    "severity": "safe",
                    "filtered": false
                }
            },
            "content_filter_offsets": {
                "end_offset": 840,
                "start_offset": 0,
                "check_offset": 0
            },
            "source_type": "prompt",
            "content_filter_raw": [],
            "blocked": false
        }
    ],
    "output": [
        {
            "phase": "final_answer",
            "role": "assistant",
            "type": "message",
            "content": [
                {
                    "annotations": [],
                    "type": "output_text",
                    "logprobs": [],
                    "text": "Of course. Please send me the specific content of 'Question A' and I will answer 'A' directly."
                }
            ],
            "id": "msg_0e27378e1f3e955e0069d4d68f1b448193870af25f1aacd3e2",
            "status": "completed"
        }
    ],
    "top_p": 0.98,
    "completed_at": 1775556239,
    "frequency_penalty": 0.0,
    "parallel_tool_calls": true,
    "background": false,
    "temperature": 1.0,
    "tool_choice": "auto",
    "model": "MaaS_GP_5.4_mini_20260317",
    "service_tier": "auto",
    "id": "resp_0e27378e1f3e955e0069d4d68ed438819387245017afa12e6c",
    "text": {
        "format": {
            "type": "text"
        },
        "verbosity": "medium"
    },
    "truncation": "disabled",
    "object": "response",
    "status": "completed"
}

/responses: Streaming Synchronous Request

Example of Curl Request

curl --location --request POST 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses' \
--header 'Authorization: Bearer {YOUR_API_KEY}' \
--header 'Content-Type: application/json' \
--data-raw '{
    "model": "gpt-5.4-mini",
    "input": "Question A: Answer A.",
    "stream": true
  }'

Python Request Example

import requests
import json

url = "https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses"

headers = {
    "Content-Type": "application/json",
    "Authorization": "Bearer {YOUR_API_KEY}"
}

payload = {
    "input": [
        {
            "role": "developer",
            "content": "Talk like a pirate."
        },
        {
            "role": "user",
            "content": "Are semicolons optional in JavaScript?"
        }
    ],
    "stream": True
}

# Use stream=True parameter to enable streaming response
with requests.post(url, headers=headers, json=payload, stream=True) as response:
    response.raise_for_status()

    # Read streaming response line by line
    for line in response.iter_lines(decode_unicode=True):
        if line:
            # Process data in SSE (Server-Sent Events) format
            if line.startswith('data: '):
                data_str = line[6:]  # Remove 'data: ' prefix
                if data_str != '[DONE]':
                    try:
                        data = json.loads(data_str)
                        print(data)
                    except json.JSONDecodeError:
                        print(data_str)

/responses: Non-streaming asynchronous request

Example of Curl Request

curl --location --request POST 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses' \
--header 'Authorization: Bearer {YOUR_API_KEY}' \
--header 'Content-Type: application/json' \
--data-raw '{
    "background": true,
    "model": "gpt-5.4-mini",
    "input": "Question A: Answer A."
  }'

Python Request Example

import requests

url = "https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses"
headers = {
    "Content-Type": "application/json",
    "Authorization": "Bearer {YOUR_API_KEY}"
}
payload = {
    "background": True,
    "input": [
        {"role": "developer", "content": "Talk like a pirate."},
        {"role": "user", "content": "Are semicolons optional in JavaScript?"}
    ]
}

response = requests.post(url, headers=headers, json=payload)
print(response.status_code)
print(response.json())

Return Example

{
    "top_logprobs": 0,
    "metadata": {},
    "presence_penalty": 0.0,
    "reasoning": {
        "effort": "medium"
    },
    "created_at": 1775617385,
    "store": true,
    "tools": [],
    "output": [],
    "top_p": 1.0,
    "frequency_penalty": 0.0,
    "parallel_tool_calls": true,
    "background": true,
    "temperature": 1.0,
    "tool_choice": "auto",
    "model": "MaaS_GP_5_mini_20250807",
    "service_tier": "auto",
    "id": "resp_04040303cf238c130069d5c569cbcc81959f0a59b62453cba4",
    "text": {
        "format": {
            "type": "text"
        },
        "verbosity": "medium"
    },
    "truncation": "disabled",
    "object": "response",
    "status": "queued"
}

/responses: Streaming asynchronous request

curl "https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer {YOUR_API_KEY}" \
    -d '{
        "background": true,
        "stream": true,
        "input": [
            {
                "role": "developer",
                "content": "Talk like a pirate."
            },
            {
                "role": "user",
                "content": "Are semicolons optional in JavaScript?"
            }
        ]
    }'

Python Request Example

import requests

url = "https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses"
headers = {
    "Content-Type": "application/json",
    "Authorization": "Bearer {YOUR_API_KEY}"
}
payload = {
    "background": True,
    "stream": True,
    "input": [
        {"role": "developer", "content": "Talk like a pirate."},
        {"role": "user", "content": "Are semicolons optional in JavaScript?"}
    ]
}

# Streaming asynchronous request
with requests.post(url, headers=headers, json=payload, stream=True) as response:
    for line in response.iter_lines(decode_unicode=True):
        if line:
            print(line)

/responses: Retrieve asynchronous request results

curl --location --request GET 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses/resp_0bd529239ed0ff590069bb8f70a8448193976feb98de696212' \
--header 'Authorization: Bearer {YOUR_API_KEY}' \
--header 'Content-Type: application/json' \
--data-raw '{}'

Return Example

{
    "top_logprobs": 0,
    "metadata": {},
    "presence_penalty": 0.0,
    "reasoning": {
        "effort": "none"
    },
    "usage": {
        "input_tokens_details": {
            "cached_tokens": 0
        },
        "input_tokens": 24,
        "output_tokens_details": {
            "reasoning_tokens": 0
        },
        "total_tokens": 233,
        "output_tokens": 209
    },
    "created_at": 1775627820,
    "store": true,
    "tools": [],
    "content_filters": [
        {
            "content_filter_results": {
                "self_harm": {
                    "severity": "safe",
                    "filtered": false
                },
                "jailbreak": {
                    "filtered": false,
                    "detected": false
                },
                "hate": {
                    "severity": "safe",
                    "filtered": false
                },
                "sexual": {
                    "severity": "safe",
                    "filtered": false
                },
                "violence": {
                    "severity": "safe",
                    "filtered": false
                }
            },
            "content_filter_offsets": {
                "end_offset": 870,
                "start_offset": 0,
                "check_offset": 0
            },
            "source_type": "prompt",
            "content_filter_raw": [],
            "blocked": false
        }
    ],
    "output": [
        {
            "phase": "final_answer",
            "role": "assistant",
            "type": "message",
            "content": [
                {
                    "annotations": [],
                    "type": "output_text",
                    "logprobs": [],
                    "text": "Arrr, aye — semicolons be **optional** in JavaScript much of the time, because the language has **automatic semicolon insertion**.\n\nBut beware:\n\n- JavaScript will **sometimes add them for ye**\n- and sometimes **not the way ye expect**\n- so omitting them can lead to weird bugs\n\n### Example\n```js\nconst a = 1\nconst b = 2\nconsole.log(a + b)\n```\nThis usually works fine.\n\n### But this can break\n```js\nreturn\n{\n  ok: true\n}\n```\nJavaScript treats that like:\n```js\nreturn;\n{\n  ok: true\n}\n```\nSo it returns `undefined`.\n\n### Rule of thumb\n- **Yes, semicolons are optional**\n- **No, it’s not always safe to skip them**\n- Many crews use semicolons anyway to avoid trouble\n\nIf ye want, I can show ye the main JavaScript gotchas with semicolon insertion, matey."
                }
            ],
            "id": "msg_03796fc2ac85b8190069d5ee30dbf88195898413def497ddd0",
            "status": "completed"
        }
    ],
    "top_p": 0.98,
    "completed_at": 1775627826,
    "frequency_penalty": 0.0,
    "parallel_tool_calls": true,
    "background": true,
    "temperature": 1.0,
    "tool_choice": "auto",
    "model": "MaaS_GP_5.4_mini_20260317",
    "service_tier": "auto",
    "id": "resp_03796fc2ac85b8190069d5ee2c38b4819582ae06f00a1817d8",
    "text": {
        "format": {
            "type": "text"
        },
        "verbosity": "medium"
    },
    "truncation": "disabled",
    "object": "response",
    "status": "completed"
}

/chat/completions

/chat/completions Non-streaming Request

curl "https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/chat/completions" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer {YOUR_API_KEY}" \
    -d '{
        "messages": [
            {
                "role": "developer",
                "content": "Talk like a pirate."
            },
            {
                "role": "user",
                "content": "Are semicolons optional in JavaScript?"
            }
        ]
    }'

Python Request Example

import requests

# API endpoint (replace with your actual endpointPath)
url = "https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/chat/completions"

# Request headers
headers = {
    "Content-Type": "application/json",
    "Authorization": "Bearer {YOUR_API_KEY}"
}

data = {
    "messages": [
        {
            "role": "developer",
            "content": "Talk like a pirate."
        },
        {
            "role": "user",
            "content": "Are semicolons optional in JavaScript?"
        }
    ]
}

# Send POST request
response = requests.post(url, headers=headers, json=data)

# Print results
print("Status code:", response.status_code)
print("Response:", response.json())

Response Example

{
    "id": "chatcmpl-DRvzyVchkK7ZR0WqwMsHUomRujwSL",
    "choices": [
        {
            "index": 0,
            "logprobs": null,
            "message": {
                "role": "assistant",
                "content": "Aye, mostly — but not always.\n\nJavaScript has a feature called **Automatic Semicolon Insertion (ASI)**, which means the engine can often add semicolons for ye when they’re omitted.\n\nExample:\n\n```js\nlet x = 5\nlet y = 10\nconsole.log(x + y)\n```\n\nThat usually works fine.\n\nBut there be **dangerous cases** where leaving them out can change the meaning or break the code. For example:\n\n```js\nlet a = 1\nlet b = 2\n[a, b].forEach(console.log)\n```\n\nJavaScript might treat that `[` as continuing the previous line, which can cause trouble.\n\nAnother classic trap:\n\n```js\nreturn\n{\n  name: \"Jack\"\n}\n```\n\nThis becomes:\n\n```js\nreturn;\n{\n  name: \"Jack\"\n}\n```\n\nSo it returns `undefined`, not the object.\n\n## Short answer\n- **Yes**, semicolons are often optional.\n- **No**, they are not always safe to omit.\n\n## Best practice\nMany crews choose one of these:\n- **Always use semicolons** for safety and clarity, or\n- **Omit them consistently** only if ye understand ASI rules well and use a formatter/linter like **Prettier** or **ESLint**.\n\nSo: **optional by syntax in many cases, but not truly optional in practice unless ye be careful.**",
                "refusal": null,
                "annotations": [],
                "images": null,
                "reasoning_content": null,
                "function_call": null,
                "tool_calls": null,
                "reasoning_details": null
            },
            "finish_reason": "stop",
            "native_finish_reason": null
        }
    ],
    "logprobs": null,
    "created": 1775550174,
    "model": "MaaS_GP_5.4_20260305",
    "object": "chat.completion",
    "system_fingerprint": null,
    "service_tier": null,
    "usage": {
        "prompt_tokens": 24,
        "completion_tokens": 299,
        "total_tokens": 323,
        "completion_tokens_details": {
            "accepted_prediction_tokens": 0,
            "audio_tokens": 0,
            "image_tokens": 0,
            "reasoning_tokens": 0,
            "rejected_prediction_tokens": 0
        },
        "prompt_tokens_details": {
            "audio_tokens": 0,
            "cached_tokens": 0
        },
        "cache_creation_input_tokens": null,
        "cache_creation": null,
        "gemini_cache_tokens_details": null
    }
}

/chat/completions Streaming Request

curl "https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/chat/completions" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer {YOUR_API_KEY}" \
    -d '{
        "messages": [
            {
                "role": "developer",
                "content": "Talk like a pirate."
            },
            {
                "role": "user",
                "content": "Are semicolons optional in JavaScript?"
            }
        ],
        "stream": true
    }'

Tool Search - Tool_search

tool_search does not support/chat/completions, only supports /responses

Reference Document: https://developers.openai.com/api/docs/guides/tools-tool-search

Request Example

curl --location --request POST 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses' \
--header 'Authorization: Bearer {YOUR_API_KEY}' \
--header 'Content-Type: application/json' \
--data-raw '{
    "input": "List open orders for customer CUST-12345.",
    "parallel_tool_calls": false,
    "tools": [
      {
        "type": "namespace",
        "name": "crm",
        "description": "CRM tools for customer lookup and order management.",
        "tools": [
          {
            "type": "function",
            "name": "get_customer_profile",
            "description": "Fetch a customer profile by customer ID.",
            "parameters": {
              "type": "object",
              "properties": {
                "customer_id": { "type": "string" }
              },
              "required": ["customer_id"],
              "additionalProperties": false
            }
          },
          {
            "type": "function",
            "name": "list_open_orders",
            "description": "List open orders for a customer ID.",
            "defer_loading": true,
            "parameters": {
              "type": "object",
              "properties": {
                "customer_id": { "type": "string" }
              },
              "required": ["customer_id"],
              "additionalProperties": false
            }
          }
        ]
      },
      { "type": "tool_search" }
    ]
  }'

Expected to see something like this in the output:

tool_search_call
tool_search_output
Subsequent function_call (e.g., list_open_orders)

Return Example

{
    "top_logprobs": 0,
    "metadata": {},
    "presence_penalty": 0.0,
    "reasoning": {
        "effort": "none"
    },
    "usage": {
        "input_tokens_details": {
            "cached_tokens": 0
        },
        "input_tokens": 593,
        "output_tokens_details": {
            "reasoning_tokens": 19
        },
        "total_tokens": 636,
        "output_tokens": 43
    },
    "created_at": 1775627905,
    "store": true,
    "tools": [
        {
            "type": "tool_search"
        },
        {
            "description": "CRM tools for customer lookup and order management.",
            "type": "namespace",
            "tools": [
                {
                    "description": "Fetch a customer profile by customer ID.",
                    "type": "function",
                    "name": "get_customer_profile",
                    "strict": true,
                    "parameters": {
                        "type": "object",
                        "required": [
                            "customer_id"
                        ],
                        "additionalProperties": false,
                        "properties": {
                            "customer_id": {
                                "type": "string"
                            }
                        }
                    }
                },
                {
                    "description": "List open orders for a customer ID.",
                    "type": "function",
                    "defer_loading": true,
                    "name": "list_open_orders",
                    "strict": true,
                    "parameters": {
                        "type": "object",
                        "required": [
                            "customer_id"
                        ],
                        "additionalProperties": false,
                        "properties": {
                            "customer_id": {
                                "type": "string"
                            }
                        }
                    }
                }
            ],
            "name": "crm"
        }
    ],
    "content_filters": [
        {
            "content_filter_results": {
                "self_harm": {
                    "severity": "safe",
                    "filtered": false
                },
                "jailbreak": {
                    "filtered": false,
                    "detected": false
                },
                "hate": {
                    "severity": "safe",
                    "filtered": false
                },
                "sexual": {
                    "severity": "safe",
                    "filtered": false
                },
                "violence": {
                    "severity": "safe",
                    "filtered": false
                }
            },
            "content_filter_offsets": {
                "end_offset": 2559,
                "start_offset": 0,
                "check_offset": 0
            },
            "source_type": "prompt",
            "content_filter_raw": [],
            "blocked": false
        }
    ],
    "output": [
        {
            "execution": "server",
            "type": "tool_search_call",
            "arguments": {
                "paths": [
                    "crm"
                ]
            },
            "id": "tsc_04a644e35c4f876d0069d5ee82bacc8194be92a9b2487c4027",
            "status": "completed"
        },
        {
            "execution": "server",
            "type": "tool_search_output",
            "tools": [
                {
                    "description": "CRM tools for customer lookup and order management.",
                    "type": "namespace",
                    "tools": [
                        {
                            "description": "List open orders for a customer ID.",
                            "type": "function",
                            "defer_loading": true,
                            "name": "list_open_orders",
                            "strict": true,
                            "parameters": {
                                "type": "object",
                                "required": [
                                    "customer_id"
                                ],
                                "additionalProperties": false,
                                "properties": {
                                    "customer_id": {
                                        "type": "string"
                                    }
                                }
                            }
                        }
                    ],
                    "name": "crm"
                }
            ],
            "id": "tso_04a644e35c4f876d0069d5ee82c86c819491796ba0edb3faf1",
            "status": "completed"
        },
        {
            "type": "function_call",
            "call_id": "call_gIstyUK0Wj4G6n79UpI8EhT8",
            "name": "list_open_orders",
            "namespace": "crm",
            "arguments": "{\"customer_id\":\"CUST-12345\"}",
            "id": "fc_04a644e35c4f876d0069d5ee83642481948f8f7e0d73831ed7",
            "status": "completed"
        }
    ],
    "top_p": 0.98,
    "completed_at": 1775627907,
    "frequency_penalty": 0.0,
    "parallel_tool_calls": false,
    "background": false,
    "temperature": 1.0,
    "tool_choice": "auto",
    "model": "MaaS_GP_5.4_mini_20260317",
    "service_tier": "auto",
    "id": "resp_04a644e35c4f876d0069d5ee81f9d8819499ad10d4b53bf86e",
    "text": {
        "format": {
            "type": "text"
        },
        "verbosity": "medium"
    },
    "truncation": "disabled",
    "object": "response",
    "status": "completed"
}

Testing the use of GPT 5.4 and GPT 5.4 Mini computer tools

1) First request (start computer loop)
You will see computer_call in the response (usually requesting a screenshot first, or returning a batch of actions[]).

curl https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses \
  -H "Authorization: Bearer {YOUR_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "input": "Please open https://example.com in the browser and tell me the page title.",
    "tools": [
      { "type": "computer" }
    ],
    "parallel_tool_calls": false
  }'



 2) After executing actions, send back the screenshot (critical)
Assuming you have:
Executed the actions[] returned by the model
Captured the latest screen and obtained BASE64_PNG
Then continue the same session:
curl https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses \
  -H "Authorization: Bearer {YOUR_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "previous_response_id": "resp_xxx",
    "input": [
      {
        "type": "computer_call_output",
        "call_id": "call_xxx",
        "output": {
          "type": "input_image",
          "image_url": "data:image/png;base64,BASE64_PNG",
          "detail": "original"
        }
      }
    ],
    "tools": [
      { "type": "computer" }
    ],
    "parallel_tool_calls": false
  }'

3) Repeat until computer_call is no longer returned
Continue executing actions[]
Take another screenshot
Send another round of computer_call_output
Until the output becomes a plain text result/final answer

background (background mode)

Both GPT-5.4 mini and GPT-5.4 nano support the background mode

Example of background asynchronous request is as follows:

curl --location --request POST 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses' \
--header 'Authorization: Bearer {YOUR_API_KEY}' \
--header 'Content-Type: application/json' \
--data-raw '{
    "background": true,
    "model": "gpt-5.4-nano",
    "input": "Question A: Answer A."
  }'

An example of the response result is as follows:

{
    "top_logprobs": 0,
    "metadata": {},
    "presence_penalty": 0.0,
    "reasoning": {
        "effort": "none"
    },
    "created_at": 1773899632,
    "store": true,
    "tools": [],
    "output": [],
    "top_p": 0.98,
    "frequency_penalty": 0.0,
    "parallel_tool_calls": true,
    "background": true,
    "temperature": 1.0,
    "tool_choice": "auto",
    "model": "MaaS_GP_5.4_nano_20260317",
    "service_tier": "auto",
    "id": "resp_0bd529239ed0ff590069bb8f70a8448193976feb98de696212",
    "text": {
        "format": {
            "type": "text"
        },
        "verbosity": "medium"
    },
    "truncation": "disabled",
    "object": "response",
    "status": "queued"
}

An example of obtaining the result of an asynchronous request is as follows:

curl --location --request GET 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/responses/resp_0bd529239ed0ff590069bb8f70a8448193976feb98de696212' \
--header 'Authorization: Bearer {YOUR_API_KEY}' \
--header 'Content-Type: application/json' \
--data-raw '{}'

Return Example

{
    "top_logprobs": 0,
    "metadata": {},
    "presence_penalty": 0.0,
    "reasoning": {
        "effort": "none"
    },
    "usage": {
        "input_tokens_details": {
            "cached_tokens": 0
        },
        "input_tokens": 24,
        "output_tokens_details": {
            "reasoning_tokens": 0
        },
        "total_tokens": 352,
        "output_tokens": 328
    },
    "created_at": 1775629269,
    "store": true,
    "tools": [],
    "content_filters": [
        {
            "content_filter_results": {
                "self_harm": {
                    "severity": "safe",
                    "filtered": false
                },
                "jailbreak": {
                    "filtered": false,
                    "detected": false
                },
                "hate": {
                    "severity": "safe",
                    "filtered": false
                },
                "sexual": {
                    "severity": "safe",
                    "filtered": false
                },
                "violence": {
                    "severity": "safe",
                    "filtered": false
                }
            },
            "content_filter_offsets": {
                "end_offset": 870,
                "start_offset": 0,
                "check_offset": 0
            },
            "source_type": "prompt",
            "content_filter_raw": [],
            "blocked": false
        }
    ],
    "output": [
        {
            "phase": "final_answer",
            "role": "assistant",
            "type": "message",
            "content": [
                {
                    "annotations": [],
                    "type": "output_text",
                    "logprobs": [],
                    "text": "Aye, they be *optional* in JavaScript, matey—mostly.\n\n- **Semicolons (`;`) aren’t required** because JavaScript has *automatic semicolon insertion* (ASI). If the parser can figure out where a statement ends, it’ll chuck in a semicolon for ye.\n- **But they’re not optional in every case**, because ASI can’t always read yer mind—and that’s when bugs be born.\n\n### When semicolons matter (gotchas)\nIf ye skip ’em, ASI can sometimes insert them in the “wrong” spot. Two classic troublemakers:\n\n1) **Line breaks before `(`, `[` , `+`, `-`, etc.**\n```js\nreturn\n{\n  name: \"Jack\"\n}\n```\nThat can turn into something like:\n```js\nreturn; \n{\n  name: \"Jack\"\n}\n```\n\n2) **Starting a line with `(` or `[` after a statement**\n```js\nlet x = 1\n(2 + 3)\n```\nASI might treat `(2 + 3)` as a new statement, not part of the previous expression.\n\n### The practical rule o’ thumb\n- **You can omit semicolons**, especially if ye use a linter/formatter like **Prettier** and follow consistent style.\n- Still, **adding semicolons often makes code more predictable**, and many folks keep ’em for safety.\n\nIf ye tell me what style guide ye follow (Airbnb? Standard? Prettier?), I can recommend the usual approach."
                }
            ],
            "id": "msg_06b47ab625f269470069d5f3deaa88819488b5ff47f694bf40",
            "status": "completed"
        }
    ],
    "top_p": 0.98,
    "completed_at": 1775629294,
    "frequency_penalty": 0.0,
    "parallel_tool_calls": true,
    "background": true,
    "temperature": 1.0,
    "tool_choice": "auto",
    "model": "MaaS_GP_5.4_nano_20260317",
    "service_tier": "auto",
    "id": "resp_06b47ab625f269470069d5f3d5d2b8819494932d297a151aa5",
    "text": {
        "format": {
            "type": "text"
        },
        "verbosity": "medium"
    },
    "truncation": "disabled",
    "object": "response",
    "status": "completed"
}

Note: Session Persistence Policy

The GP response API involves multi-round conversations, and if there is cross-account access, there will be issues with request failures. To address this problem, you can add X-Conversation-Id to the request header, with the value being the conversation ID, and the ID for the same round of conversation needs to be consistent. This conversation persistence also has a time limit, with a maximum duration of 30 minutes.

response request

curl --location --request POST 'https://genaiapi.cloudsway.net/v1/ai/{your endpoint}/responses' \
--header 'Authorization: Bearer {YOUR_API_KEY}' \
--header 'X-Conversation-Id: conversation-test1' \
--header 'Content-Type: application/json' \
--data-raw '{
  "input": "Tell me another brain teaser with the answer. Return all of them together.",
  "previous_response_id": "resp_04825fd1c08c48820069c12ebdd7948190ac42abb8234f1a83",
  "stream": true
}'

Chat completion

curl --location --request POST 'https://genaiapi.cloudsway.net/v1/ai/{endpointPath}/chat/completions' \
--header 'Authorization: Bearer {YOUR_API_KEY}' \
--header 'X-Conversation-Id: conversation-test1' \
--header 'Content-Type: application/json' \
--data-raw '{

    "messages": [
        {
            "role":"user",
            "content":"hi"
        }
    ]
}'

SDK Call

1. Basic Usage

from openai import OpenAI

client = OpenAI(
    api_key="xxxx",
    base_url="https://genaiapi.cloudsway.net/v1/ai/xxxx/"
)

resp = client.chat.completions.create(
    model="gpt5-pro",
    messages=[{
            "role": "user",
            "content": "How much gold would it take to coat the Statue of Liberty in a 1mm layer?"
        }]
)

print(resp.choices[0].message.content)

2. Function calling

from openai import OpenAI
import json

client = OpenAI(
    api_key="xxxx",
    base_url="https://genaiapi.cloudsway.net/v1/ai/xxxx/"
)

# 1. Define a list of callable tools for the model
tools = [
    {
        "type": "function",
        "name": "get_horoscope",
        "description": "Get today's horoscope for an astrological sign.",
        "parameters": {
            "type": "object",
            "properties": {
                "sign": {
                    "type": "string",
                    "description": "An astrological sign like Taurus or Aquarius",
                },
            },
            "required": ["sign"],
        },
    },
]

def get_horoscope(sign):
    return f"{sign}: Next Tuesday you will befriend a baby otter."

# Create a running input list we will add to over time
input_list = [
    {"role": "user", "content": "What is my horoscope? I am an Aquarius."}
]

# 2. Prompt the model with tools defined
response = client.responses.create(
    model="gpt-5",
    tools=tools,
    input=input_list,
)

# Save function call outputs for subsequent requests
input_list += response.output

for item in response.output:
    if item.type == "function_call":
        if item.name == "get_horoscope":
            # 3. Execute the function logic for get_horoscope
            horoscope = get_horoscope(json.loads(item.arguments))

            # 4. Provide function call results to the model
            input_list.append({
                "type": "function_call_output",
                "call_id": item.call_id,
                "output": json.dumps({
                  "horoscope": horoscope
                })
            })

print("Final input:")
print(input_list)

response = client.responses.create(
    model="gpt-5",
    instructions="Respond only with a horoscope generated by a tool.",
    tools=tools,
    input=input_list,
)

# 5. The model should be able to give a response!
print("Final output:")
print(response.model_dump_json(indent=2))
print("\n" + response.output_text)

3. Thinking

from openai import OpenAI

client = OpenAI(
    base_url="https://genaiapi.cloudsway.net/v1/ai/xxxx",
    api_key="xxxx"
)

prompt = """
Write a bash script that takes a matrix represented as a string with 
format '[1,2],[3,4],[5,6]' and prints the transpose in the same format.
"""

response = client.responses.create(
    model="gpt-5pro",
    reasoning={"effort": "medium"},
    input=[
        {
            "role": "user", 
            "content": prompt
        }
    ]
)

print(response.output_text)

MaaS_GP_5.4

Basic Information

New Feature Introduction

Model Capability List

Request Parameters

Newphaseparameter

Request Example

/responses

/responses: Non-streaming synchronous request

Return Example

/responses: Streaming Synchronous Request

/responses: Non-streaming asynchronous request

Return Example

/responses: Streaming asynchronous request

Python Request Example

/responses: Retrieve asynchronous request results

Return Example

/chat/completions

/chat/completions Non-streaming Request

/chat/completions Streaming Request

Tool Search - Tool_search

Request Example

Return Example

Testing the use of GPT 5.4 and GPT 5.4 Mini computer tools

background (background mode)

Note: Session Persistence Policy

SDK Call

1. Basic Usage

2. Function calling

3. Thinking

New`phase`parameter