One prompt through all 19 Sume image model ids costs $1.36

One prompt on every image id in Sume's catalog at default tier: $1.36 for 19 images, $1.09 without the older ChatGPT Image 2. Table and a Python loop.

6 min readSume
All posts

If you are choosing an image model for a product, the cheapest honest test is one prompt through everything. Sume's catalog lists 19 canonical image ids today (GET /v1/images/models, read 2026-10-07; google/nano-banana-2 is retired and runs as 2.1). At each model's default tier, one image apiece costs $1.358 in total.

Drop the older openai/gpt-image-2, which is the single most expensive row at its catalog display rate of high quality at 1024 square, and the sweep is $1.09425. That is the price of looking at 18 different answers to one brief.

Price per image at default tier

Prices here are Sume's list-times-1.25 figures. The catalog states the billable formula as "list × 1.25 → ceil usd cents", so treat the dollar amounts as the pre-rounding value and read the exact charge from billable_amount_usd_micros in the submit envelope. Failed or cancelled generations are not billed.

Sume image catalog display rates, n=1, read 2026-10-07. GPT Image 2.5 prices at the high-quality 1024 estimate; Nano Banana is the 1K rate.
Model idTier pricedPrice per image
higgsfield/soul720p$0.005
x-ai/grok-imagedefault$0.025
qwen/qwen-imagedefault$0.025
google/imagen-4-fastdefault$0.025
bytedance-seed/seedream-4default$0.0325
black-forest-labs/flux.2-prodefault$0.0375
bytedance-seed/seedream-5-litedefault$0.04375
bytedance-seed/seedream-4.5default$0.05
recraft/recraft-v4default$0.05
black-forest-labs/flux.2-flexdefault$0.0625
openai/gpt-image-2.5high, 1024$0.06588
openai/gpt-image-2.5-sunbursthigh, 1024$0.06588
google/imagen-4-ultra2K default$0.075
ideogram/ideogram-v3default$0.075
ideogram/ideogram-v4.5medium$0.075
qwen/qwen-image-maxdefault$0.09375
google/nano-banana-2.11K$0.10
google/nano-banana-pro1K$0.1875
openai/gpt-image-2high, 1024$0.26375
All 19, one image each$1.358

Reading the sweep

Six of the 19 ids bill under 5 cents: Soul, Seedream 4.0, Grok Imagine, Qwen Image, Imagen 4 Fast and Seedream 5 Lite. Use them for the first look. Pay for a pricier model only on the brief where the cheap ones fail, not by default.

Five ids are text-to-image only (Soul, Imagen 4 Fast, Imagen 4 Ultra, Recraft V4, Qwen Image Max). If your brief includes a reference photo, the sweep is 14 ids, not 19: the rest return 400 on input_references.

The loop

It lists the catalog, sends the same prompt to each id, and prints the cost Sume reports. Use mode: "async" if you would rather collect results later; here we read each response and accept that some may come back as 202.

import os, requests
H = {"Authorization": "Bearer " + os.environ["SUME_API_KEY"]}
B = "https://api.sume.com/v1/images"
models = requests.get(B + "/models", headers=H, timeout=30).json()["data"]
total = 0.0
for m in models:
    r = requests.post(B, headers=H, timeout=60,
                      json={"model": m["id"], "prompt": "a ceramic teapot on a wooden table, soft window light"})
    if r.status_code == 200:
        body = r.json()
        total += body["usage"]["cost"]
        print(m["id"], body["data"][0]["url"])
    else:
        print(m["id"], r.status_code)
print("total", round(total, 4))

Source: Sume Image API docs (read 2026-10-07). The loop sends one request per catalog row and bills each completed image.

Sources

Related posts

More in Pricing

All Pricing posts

Written by Sume