MaaS Open Platform API Documentation

Domain

https://hiharness.ai

Authentication

Use the Authorization header for Bearer authentication. The API key format is sk-xxxxxxxx.

Example:

Authorization: Bearer <YOUR_API_KEY>

Basic Workflow

Image and video models use asynchronous tasks:

1. Submit a generation task and receive a task_id.
2. Poll the result endpoint with the task_id.

Large language models use /v1/chat/completions. Non-streaming requests return the full assistant message synchronously, and streaming requests return incremental content through SSE.

Available Models

Image Models

Hidream-O Series

• HiDream-O1-Image-Dev-2604 HiDream-O1-Image-Dev-2604
• HiDream-O1-Image-1.5 HiDream-O1-Image-1.5
• HiDream-O1-Image-2 HiDream-O1-Image-2
• HiDream-O1-Edit-1.5 HiDream-O1-Edit-1.5

Hidream-Q Series

• Image-2zvskglc hidream-Q3-pro-Image
• Image-f3qa3lph hidream-Q3-std-Image
• Image-aw47boi4 hidream-Q1-image

Hidream-H Series

• Image-qyoyq2bi hidream-H4.5-image
• Image-84977rs4 hidream-H4.0-image

Other Image Models

• Image-zksa5by1 hidream-tryon-v1
• Image-eccut7dd hidream-translate-v1
• Image-q5580iqf hidream-image-erase-v1
• Image-580g70sx hidream-outpainting-v1
• Image-11lqtks8 hidream-superresolution-v1

Video Models

HiDream-O Series

• HiDream-O1-Video-1.0 HiDream-O1-Video-1.0

Hidream-Q Series

• Video-6bj3i4u2 hidream-Q3-pro-Video
• Video-jkpfq9r0 hidream-Q3-std-Video
• Video-7flw327g hidream-Q2.6-Video
• Video-idegmbam hidream-Q2.5-pro-Video
• Video-3o4t26br hidream-Q1-video

Hidream-H Series

• Video-eol8ra2d hidream-H1.5-video
• Video-9plmpqyo hidream-H1.0-video

Other Video Models

• Video-t2ze92dg avatar_image2video (digital human video)

Large Language Models

• qwen3.5-plus
• qwen3.6-plus
• qwen3.7-plus
• qwen3.7-max
• kimi-k2.6
• deepseek-v3-2-251201
• deepseek-v4-flash
• deepseek-v4-pro

Sample Code

import time
import requests
import time

import requests

# Host and token
HOST = ""
TOKEN = "sk-xx"

head = {"Authorization": f"Bearer {TOKEN}"}


def send_task(body: dict, path):
    url = f"{HOST}{path}"
    # Print request parameters to help troubleshoot issues
    # print(url, body, head)
    res = requests.post(url, json=body, headers=head)
    assert (
        res.status_code == 200 and res.json().get("code") == 0
    ), f"{res.status_code}, {res.text}"
    return res.json()


def get_task_result(task_id: str, path):
    url = f"{HOST}{path}"
    res = requests.get(url, headers=head, params={"task_id": task_id})
    assert (
        res.status_code == 200 and res.json().get("code") == 0
    ), f"{res.status_code}, {res.text}"
    return res.json()


def main():
    # Note that image and video tasks use different endpoint paths
    request_path = f"/api/maas/gw/v1/images/generations"
    # Model parameters (common example; see specific model docs for details)
    request_params = {
        "model_id": "Image-xx",
        "prompt": "A beautiful landscape",
        "request_id": "my-task-001",
        "n": 1,
        "negative_prompt": "",
    }

    print(f"Starting task {request_params.get('model_id')}")
    response_json = send_task(request_params, path=request_path)
    task_id = response_json["result"]["task_id"]
    print(f"Task {task_id} created successfully!")
    result_path = request_path + "/results"
    while True:
        response_json = get_task_result(task_id, result_path)
        result = response_json["result"]
        if result["status"] == 1:
            print(f"Task {task_id} completed:", result)
            break
        else:
            print(f"Task {task_id} is still processing", result)
            time.sleep(5)


if __name__ == "__main__":
    main()