fal image model list vs the Sume catalog: how to check by query

To find out whether a model from fal's gallery is callable through Sume, do not compare web pages: query the catalog. GET /v1/images/models returns every image model your key can use, with the parameters each one accepts, so a name search takes seconds.

What fal's gallery lists

On October 3, 2026, the fal.ai model gallery lists, among others, Seedream 5.0, GPT Image 2.5, Flux 2, Nano Banana 2, Ideogram 4 and Krea 2 for images. It also describes Seedance 2.5 as producing a native 30-second clip at up to 720p, which matters if you compare video options as well.

Query the Sume catalog

The Sume Image API publishes model ids in an org/slug form, for example openai/gpt-image-2.5 and ideogram/ideogram-v4.5, and also accepts bare legacy ids such as nano-banana-2 as aliases. Because names differ between catalogs, search the id and the display name together, and treat a miss as a prompt to read the list by eye rather than proof that a model is absent.

import os
import requests

KEYWORDS = ["seedream", "gpt-image", "flux", "banana", "ideogram", "krea"]


def main() -> None:
    r = requests.get(
        "https://api.sume.com/v1/images/models",
        headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
        timeout=30,
    )
    r.raise_for_status()
    models = r.json()["data"]
    for key in KEYWORDS:
        hits = [m["id"] for m in models if key in (m["id"] + " " + m.get("name", "")).lower()]
        print(f"{key:10s} -> {hits or 'no match; read the list by eye'}")


if __name__ == "__main__":
    main()

What to compare once you find a match

Names are the start of the comparison, not the end. For each match, read the endpoint record at GET /v1/images/models/{model_id}/endpoints. It lists the supported parameters, such as how many input_references the model takes and which aspect ratios it accepts, and a pricing line that is the amount charged. A request that sets a parameter the model does not list returns 400 unsupported_parameter, so a catalog check up front saves a failed run later.

• Reference images: read the input_references range, since some models are text-to-image only.
• Aspect ratios: read the aspect_ratio enum before you pin one.
• Price: read pricing[].cost_usd and multiply by n.
• Streaming: every catalog row currently reports supports_streaming: false.

A note on video

The gallery's Seedance 2.5 description is a vendor claim about clip length and resolution. For video, check a catalog the same way and read the stated limits for your target model before you plan a 30-second clip.