OpenAI Image API vs Sume Image API for GPT Image 2.5: what differs
Same model family, different request surface. Moderation, revised_prompt, streaming, output_compression and masks compared between OpenAI direct and Sume.
Both OpenAI and Sume serve GPT Image 2.5, but the request and response surfaces differ. Sume returns hosted URLs rather than base64, takes a mask as a URL, accepts up to 16 reference photos, and does not serve streaming, output_compression or seed. If your code relies on those OpenAI fields, plan changes before you move it.
Side-by-side
OpenAI facts are from its image generation guide; Sume facts are from the Image API docs.
| Feature | OpenAI direct | Sume |
|---|---|---|
| Model ids | gpt-image-2.5-sunburst, gpt-image-2.5-flare | openai/gpt-image-2.5-sunburst, openai/gpt-image-2.5 (Flare) |
| Quality | low, medium, high, xhigh, max, auto | Same values; omitted means high |
| Mask | File with alpha channel, same size | mask_url, public HTTPS |
| Streaming, partial_images 0 to 3 | Supported | Not served; stream true returns 400 |
| output_compression | 0 to 100 for jpeg and webp | Not served; returns 400 |
| moderation auto or low | Supported | Not listed, so rejected |
| revised_prompt | Returned | Not in the response |
| Result | Image data | Sume-hosted signed URL in data[].url |
| Wait | Up to about 2 minutes for complex prompts | 30 seconds, then a 202 job |
What Sume adds
Sume gives you one API key and one job model across images, video and audio. Billing is all-or-nothing: failed or cancelled generations are not charged. Reference photos go up to 16 on the 2.5 rows, while the OpenAI guide shows examples with up to 4. Idempotency-Key and webhook mode follow the standard Sume job rules in Jobs and results.
When to stay direct
If you depend on partial image streaming, jpeg compression control or the low moderation setting, call OpenAI directly for those paths. Use Sume when one job model, hosted URLs and a mixed catalog matter more than those fields.
Sources
Related posts
More in Comparisons
- OpenRouter video client on Sume: six documented differences
Sume says /v1/videos matches the OpenRouter Video Generation API except six items: base path, auth, model ids, auto routing, size and provider.options.
- Pick an image model by job: product photo, poster text, edit
Five new image releases in a month. A job-first table for Sume: GPT Image 2.5, Ideogram 4.5, Nano Banana 2.1, FLUX.2 and Qwen, with the limits that decide.
- Sora API shutdown report: which claims Sume docs can answer
Pondero's Sora API shutdown report names Runway, Kling and Seedance. Sort its claims from what Sume's docs list, so you do not assume a model is there.
- Precise image editing on Sume: what it has and what it lacks
A plain list of what Sume offers for precise edits after Flux 3 Image and Ideogram 4.5: no Flux 3, no box field, mask_url on one model, Ideogram 4.5 present.
Written by Sume