Hy Image 3.5 lists a 1.5K tier. No Sume image model does

Hy Image 3.5 Preview offers four resolution tiers: 1K, 1.5K, 2K and 4K, per its OpenRouter page (read 2026-10-08). No image row in Sume's catalog lists a 1.5K tier. The tier lists on Sume are 512, 1K, 2K and 4K on the Nano Banana rows, 1K and 2K on Imagen 4 Ultra and Ideogram 4.5, and 720p and 1080p on Higgsfield Soul. To get something close to 1.5K, send image_size 1536x1536 to a GPT Image 2.5 row. That is 2,359,296 pixels, and it is legal.

Tiers on Sume today

Only some rows have a resolution descriptor at all. The others take their size from aspect_ratio or from image_size. The catalog normalizes the half-K tier to 512, and a request may send either 512 or 0.5K.

The OpenRouter page notes that concrete pixel dimensions for the Hy tiers are derived per provider, so we do not claim an exact pixel size for its 1.5K tier.

A 1.5K-class size on GPT Image 2.5

GPT custom sizes need both edges as multiples of 16, a longest edge of at most 3840, a ratio of at most 3:1 and 655,360 to 8,294,400 pixels. 1536 is 96 x 16. So 1536x1536, 1536x1024 and 1536x864 all pass. 1536x864 is a 16:9 frame at 1,327,104 pixels.

Pass the size as image_size. It wins over aspect_ratio, so do not send both and expect a blend.

{
  "model": "openai/gpt-image-2.5",
  "prompt": "A harbor at dawn, wide establishing shot",
  "image_size": {"width": 1536, "height": 864},
  "quality": "medium"
}

If you need the tier name itself

A tier name is a convenience, not a feature. If your downstream code expects a '1.5K' option, map it to a pixel size on the GPT rows, or pick the 1K or 2K tier on a Nano Banana row and resample. GPT Image 2.5 is billed from image tokens, so a bigger size costs more than a smaller one at the same quality. Read the price line of the endpoint before you batch.

Sume does not list Hy Image 3.5 Preview, so there is no Sume tier to compare with directly. Check GET /v1/images/models for the current list, since rows are added and retired.

If you need a size between the 1K and 2K tiers on a Sume row, use image_size on a custom-pixel row instead of a tier. GPT Image 2.5, Seedream, Flux, Qwen and Recraft accept width and height within their rules, so you can ask for an exact in-between size and skip the tier. Read the row's supported_parameters first to see whether it lists resolution, image_size or only aspect_ratio, because the three are not interchangeable and a field a row does not list returns 400 unsupported_parameter.