MiniMax H3 Image API

The official API exposes MiniMax-H3-Image through two synchronous, OpenAI-style endpoints. Both models, MiniMax-H3-Image and MiniMax-H3-Image-Highspeed, take the same parameters and limits.

Base URL: https://api.minimax.io (global) or https://api.minimax.cn (mainland China). Authenticate with Authorization: Bearer <API key>; keys are created in the MiniMax platform console.

Text-to-image with curl

Request body taken from the official API reference example.

curl --request POST \
  --url https://api.minimax.io/v1/images/generations \
  --header "Authorization: Bearer $MINIMAX_API_KEY" \
  --header 'Content-Type: application/json' \
  --max-time 600 \
  --data '{
    "model": "MiniMax-H3-Image",
    "prompt": "An orange cat lying on a windowsill basking in the sun, warm afternoon light, film texture",
    "size": "2720x1536",
    "n": 1,
    "response_format": "url"
  }'

Text-to-image with Python (requests)

import os
import requests

url = "https://api.minimax.io/v1/images/generations"
api_key = os.environ.get("MINIMAX_API_KEY")
headers = {"Authorization": f"Bearer {api_key}"}

payload = {
    "model": "MiniMax-H3-Image",
    "prompt": "An orange cat lying on a windowsill basking in the sun, warm afternoon light, film texture",
    "size": "2720x1536",
    "n": 1,
    "response_format": "url",
}

response = requests.post(url, headers=headers, json=payload, timeout=600)
response.raise_for_status()

result = response.json()
ext = result["output_format"]

# Download the images and store them yourself
for i, item in enumerate(result["data"]):
    if "error" in item:
        print(f"Image {i} failed: {item['error']['message']}")
        continue
    with open(f"output-{i}.{ext}", "wb") as f:
        f.write(requests.get(item["url"]).content)

Text-to-image with the OpenAI SDK

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.minimax.io/v1",
    api_key=os.environ.get("MINIMAX_API_KEY"),
    timeout=600,
    max_retries=0,  # not idempotent: disable automatic retries
)

result = client.images.generate(
    model="MiniMax-H3-Image",
    prompt="An orange cat lying on a windowsill basking in the sun, warm afternoon light, film texture",
    size="2720x1536",
    response_format="url",
)

for i, item in enumerate(result.data):
    print(i, item.url)  # url is None for images that failed

Image edit (multiple references)

Reference images go in images[].image_url as a public URL, an uploaded mm_file://{file_id}, or a data:image/...;base64, URI. Their order is the picture number the prompt refers to. Edits always run prompt enhancement.

curl --request POST \
  --url https://api.minimax.io/v1/images/edits \
  --header "Authorization: Bearer $MINIMAX_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "MiniMax-H3-Image",
  "prompt": "Replace the cat in image 1 with the Shiba Inu in image 2, keep the background unchanged",
  "images": [
    {"image_url": "https://example.com/cat.jpg"},
    {"image_url": "https://example.com/dog.png"}
  ],
  "size": "auto",
  "response_format": "url"
}'

result = client.images.edit(
    model="MiniMax-H3-Image",
    image=[open("cat.jpg", "rb"), open("dog.png", "rb")],
    prompt="Replace the cat in image 1 with the Shiba Inu in image 2, keep the background unchanged",
    size="auto",
    response_format="url",
)

Reference image limits: JPG, JPEG, PNG, WEBP, HEIC or HEIF; each file ≤ 64 MiB; width and height 128–8192 px; long/short side ≤ 4:1; 1 to 16 images; whole request body ≤ 64 MiB. Masks are not supported.

Key parameters

Common sizes: 2K 16:9 is 2720x1536, 4K 1:1 is 4096x4096, 4K 16:9 is 5472x3072, 4K 3:1 is 7104x2368.

Errors worth handling

• HTTP 200 can still contain failed items. Check data[].error on every image.
• 1026: the enhanced prompt failed the content safety check. 1027: the output image failed it.
• 402 with 1008: insufficient balance. 429 with 1002: images-per-minute limit, where each request counts as n images.

Pricing

Billing is per token, counting successful images only. Prices below are from MiniMax's pricing page, read on 2026-11-11.

MiniMax's example cost per image (Standard vs Highspeed)

MiniMax notes these are rounded reference examples, not fixed per-image prices; some 1K/2K amounts were adjusted for presentation.

China platform (CNY), per MiniMax's announcement

MiniMax's WeChat launch article lists reference costs per image on the China platform (api.minimax.cn):

Speed: MiniMax's published latency

MiniMax's launch article includes a chart of median (P50) generation time in seconds. These are MiniMax's figures and have not been independently verified.

The same chart lists other models (Nano Banana 2.1, GPT-image 2.5/2.6, Seedream 5 Pro) for comparison; see the original article.