Which aspect ratios does each Sume image model accept? Full matrix

Every Sume image model publishes its own aspect_ratio list, and a ratio that is not on it is rejected with 400 unsupported_parameter. The matrix below was built from the catalog descriptors on 2026-10-03. The normalized vocabulary on the Image API page is wider than any single model, so 16:9 and 1:1 are safe nearly everywhere, and everything else needs a look.

Three patterns stand out. 3:1 and 1:3 appear only on the two Ideogram rows. 4:1, 1:4, 8:1 and 1:8 appear only on Nano Banana 2. And the Grok Imagine row uses its own phone-shaped ratios, 20:9, 19.5:9, 9:19.5 and 9:20, which no other row lists.

The matrix

yes means the model lists the ratio. The auto column is separate and follows the table: ChatGPT Image 2, ChatGPT Image 2.5, ChatGPT Image 2.5 Sunburst, Nano Banana 2, Nano Banana Pro, Seedream 4.0 list auto.

Picking by ratio

• 1:1 and 3:4: 19 and 19 of the 19 listed ids.
• 4:5 and 5:4 are missing on: Soul (Higgsfield), Grok Imagine, Imagen 4 Fast, Imagen 4 Ultra.
• 2:3 is missing on: ChatGPT Image 2, ChatGPT Image 2.5, ChatGPT Image 2.5 Sunburst, Imagen 4 Fast, Imagen 4 Ultra.
• Ultrawide 21:9: Nano Banana 2, Nano Banana Pro, Qwen Image, Qwen Image Max, Flux 2 Pro, Flux 2 Flex, Recraft V4.
• Narrow banner 3:1: Ideogram V3, Ideogram 4.5. Extreme banners 4:1 and 8:1: Nano Banana 2.

Check in code before you send

The table is a snapshot; this loop reads the live list for the ratio you need and prints the rows that accept it.

import os
import requests

headers = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
models = requests.get("https://api.sume.com/v1/images/models", headers=headers, timeout=30).json()["data"]
ratio = "21:9"
for m in models:
    if ratio in m["supported_parameters"]["aspect_ratio"]["values"]:
        print(m["id"])