Seedream 5.0 Pro API: Sume lists 5.0 Lite, not Pro

Sume does not list a Seedream 5.0 Pro model. As of 2026-09-29 the catalog has three Seedream ids: bytedance-seed/seedream-5-lite, bytedance-seed/seedream-4.5 and bytedance-seed/seedream-4. For the newest of them, call seedream-5-lite. This post says nothing about a Pro release itself, only about what Sume serves.

Which Seedream ids can I call today?

All three take up to 10 reference images and 1 to 4 images per request. The prices are Sume's per-image estimates, rounded up to a cent, and the endpoint record shows the exact figure.

What happens if I send a Pro model id?

Any id outside the catalog returns 404 model_not_found with the message Unknown image model id. before anything is generated, so nothing is billed. Sume's docs give no Pro id to try, so guessing one is not useful.

How do I call Seedream 5.0 Lite?

Send model and prompt to POST /v1/images. aspect_ratio takes one of 9 listed ratios, including 4:5, and auto is not one of them. It does not list resolution or quality, so sending either returns 400 unsupported_parameter.

curl -X POST "https://api.sume.com/v1/images" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bytedance-seed/seedream-5-lite",
    "prompt": "a red panda astronaut floating in space, studio lighting",
    "aspect_ratio": "4:5"
  }'

How will I know if a Pro id is added?

It would appear in GET /v1/images/models with its own parameters and price; the Image API docs are the place to read how to interpret those descriptors. The docs make no promise either way, and neither does this post. For the full Seedream call shapes, see Seedream 4.5 API on Sume.

How is a Seedream call billed?

Per output image. GET /v1/images/models/bytedance-seed/seedream-5-lite/endpoints returns a pricing line for output_image, and Sume's margin is already applied, so cost_usd times n is what you pay. Sume computes its per-image estimates from a list price and rounds up to the cent, which is why the table's figures are whole cents.

A completed generation is billed in full. A failed or cancelled one is not billed, and the synchronous path returns 502 Bad Gateway when generation fails. A request slower than 30 seconds returns 202 with a job to poll instead of an image.

Should I use 4.5, 4.0 or 5.0 Lite?

Sume's docs do not rank the three, so this post does not either. What the catalog shows is narrower: 4.5 and 5.0 Lite list the same aspect ratios, and 4.0 lists those plus auto, which suits an edit that should match its reference. It is also the lowest-priced of the three in the table. Run the same prompt on two of them and compare the images; a failed generation costs nothing.