Hy-Image-3.5-Preview API limits vs Sume's image limits
Tencent's Hy-Image-3.5-Preview takes 256 to 8192 px edges, up to 4096x4096 and 20 references. Sume lists no Hy Image model; its own limits are below.
Tencent Cloud's model hy-image-v3.5-preview takes width and height from 256 to 8192 with an area of at most 16,777,216 (4096x4096), and up to 20 reference images of at most 20 MB each. Sume lists no Hy Image model, so none of these limits apply there; Sume enforces its own, below.
Tencent figures are from its guide; Sume figures from the Image API docs and catalog code. Read 2026-09-30.
What does Tencent publish for the 3.5 Preview?
It uses the Chat/Messages protocol, accepts multimodal input and multi-turn editing context, and runs synchronously. References can be public http(s) URLs or base64 data URLs, and the guide says the server truncates by round from farthest to nearest when the limit is exceeded. The earlier hy-image-v3 is a different entry; see HunyuanImage 3.0.
How do those limits compare with Sume's?
| Limit | Hy-Image-3.5-Preview (Tencent) | Sume |
|---|---|---|
| Reference images | Up to 20, each up to 20 MB | Up to 16 on GPT Image 2.5; 10 is the default ceiling for other edit-capable models |
| Custom size | Edges 256 to 8192, area up to 16,777,216 | GPT Image 2.5: edges multiples of 16, maximum edge 3840, aspect ratio at most 3:1, 655,360 to 8,294,400 pixels |
| Reference source | http(s) URL or base64 | Public HTTPS URL only |
| Long calls | Synchronous | POST /v1/images blocks up to 30 seconds, then returns a 202 job |
What happens if I send Sume a 4096x4096 request?
On GPT Image 2.5 the largest edge is 3840, so 4096 on either side is outside the documented rule. Check the endpoint's supported_parameters: Sume serves every catalog model through one sume endpoint in v1, and a value the model does not list is refused with 400 unsupported_parameter.
Where do I start on Sume?
Read the catalog entry for the model you want, then follow GPT Image 2.5 custom sizes or the reference image rules. If a call runs past 30 seconds, poll the job instead of resending.
Sources
Related posts
More in Models
- Ideogram 4.5 Precise Edit API: mask and reference limits
Ideogram 4.5 Precise Edit takes 4 reference images, 3 when a mask is sent. Sume lists Ideogram V3, and mask_url there is only for ChatGPT Image 2.5.
- How many reference images? 10 on Image 1.0, 16 on GPT Image 2.5
Image 1.0 image_urls takes 1 to 10 public HTTPS URLs; ChatGPT Image 2.5 takes up to 16 references. Text-only models reject references.
- Image quality default on Sume: low on Image 1.0, high on GPT 2.5
Image 1.0 defaults quality to low; ChatGPT Image 2.5 on POST /v1/images defaults to high when omitted. Values, and when to raise quality.
- gpt-image-2 or openai/gpt-image-2? Model ids on Sume's Image API
Sume accepts bare Image Router ids such as gpt-image-2 and nano-banana-2 as aliases for openai/gpt-image-2 style ids. The catalog lists org/slug ids.
Written by Sume