gpt-image-1-mini replacement on Sume: what each image id costs

gpt-image-1-mini shuts down December 1, 2026. Compare billed Sume prices per image for budget ids and see why mini token rates don't map one to one.

5 min readSume
All posts

If gpt-image-1-mini was your cheap tier, Sume's budget image ids bill between $0.03 and $0.05 per image for the ones listed below. OpenAI lists gpt-image-1-mini for shutdown on December 1, 2026, with gpt-image-2.5-sunburst or gpt-image-2.5-flare as the named replacement. Those are not budget models, so a cheap tier means picking a different family.

OpenAI prices mini by tokens: $2.50 per million image input tokens and $2.00 per million text input tokens on the pricing page read 2026-10-06. Sume prices per image, so a straight conversion needs your own render sizes.

Sume list and billed price per image

Treat the billed column as an estimate and read usage.cost on a real response for the exact charge.

Sume list prices x 1.25, rounded up to the cent; Sume provider-pricing tables, read 2026-10-06
Sume idList USDBilled USD (est.)
x-ai/grok-image0.020.03
qwen/qwen-image0.020.03
google/imagen-4-fast0.020.03
black-forest-labs/flux.2-pro0.030.04
ideogram/ideogram-v4.5 (low)0.030.04
bytedance-seed/seedream-5-lite0.0350.05
recraft/recraft-v40.040.05
google/nano-banana-20.080.10

Compute the billed price yourself

Use Decimal so rounding up to the cent is exact. Floats turn 0.035 x 1.25 into a value that can land on the wrong side of a cent.

from decimal import Decimal, ROUND_CEILING

LIST = {
    "x-ai/grok-image": "0.02",
    "black-forest-labs/flux.2-pro": "0.03",
    "bytedance-seed/seedream-5-lite": "0.035",
    "google/nano-banana-2": "0.08",
}

def billed(list_usd):
    raw = Decimal(list_usd) * Decimal("1.25")
    return raw.quantize(Decimal("0.01"), rounding=ROUND_CEILING)

if __name__ == "__main__":
    for model, price in LIST.items():
        print(f"{model:34} {billed(price)}")

What you give up by going cheaper

The budget ids differ from mini in prompt style, text rendering and edit support. Render your own 20 golden prompts on two candidates before you move traffic, and keep the new id in config so the model map stays one file. For the ChatGPT-family replacement and its price rules, see the 1.5 and mini post.

Sources

Related posts

More in Pricing

All Pricing posts

Written by Sume