Five fall 2026 image launches: which can an app call on Sume today?
GPT Image 2.5, FLUX 3 Image, Nano Banana 2, Midjourney V8.2 Edit and Ideogram 4.0 against the Sume catalog: two are callable by id, three are not.
Each launch has a headline feature. A buyer's question is smaller: can I call it from my app today, and through what? This post lines up five launches against the Sume image catalog as of 2026-10-04 and says plainly where there is no id to call.
The five, side by side
Sources: OpenAI Developer Community, Black Forest Labs docs, Google AI for Developers, Midjourney updates feed, Ideogram and the Sume Image API.
| Launch | Vendor fact | Callable on Sume by id? |
|---|---|---|
| ChatGPT Images 2.5 (Sept 8, 2026) | Sketch via @Sketch in ChatGPT; Flare and Sunburst API variants; quality up to xhigh and max | Yes: openai/gpt-image-2.5 and openai/gpt-image-2.5-sunburst |
| FLUX 3 Image | Boxes on a 0 to 1000 grid; up to 10 references; native 2K and 4K | No: the catalog has black-forest-labs/flux.2-pro and flux.2-flex |
| Nano Banana 2 | gemini-3.1-flash-image; up to 14 references in total; 512 px to 4K; search grounding | Yes: google/nano-banana-2 (no grounding field) |
| Midjourney V8.2 Edit (tests from Aug 27, 2026) | Text edits, up to 4 references, inpaint and outpaint | No Midjourney row in the catalog |
| Ideogram 4.0 (June 3, 2026) | 9.3B open weights, 256 to 2048 px, 0.97 OCR claim | No: the catalog has ideogram/ideogram-v4.5 and ideogram-v3 |
What that means for an edit workflow
Check GET /v1/images/models before you plan: rows are added and removed, and this table is only a dated snapshot.
- Reference-based edits and masked edits are available now: ChatGPT Image 2.5 takes up to 16 references and
mask_url. - Box-style layout control has no Sume field; describe regions in words or composite back.
- Grounded generation has no Sume field; fetch your own references.
- Open weights are a hosting choice. Ideogram's licensing page lists a free non-commercial path and a $300 per month self-serve commercial tier starting at 10,000 images a month (Ideogram licensing, read 2026-10-04).
A first test for each
Pick one real photo and one edit instruction. Run it on the two models you can call, log cost and result, and decide with a blind review. For the three you cannot call, the honest answer today is to use the nearest catalog model or check back after the catalog changes.
Sources
- OpenAI Developer Community: Introducing ChatGPT Images 2.5 (read 2026-10-04)
- Black Forest Labs docs: FLUX 3 Image editing (read 2026-10-04)
- Google AI for Developers: image generation with Gemini (read 2026-10-04)
- Midjourney updates feed: V8.2 and Edit model posts (read 2026-10-04)
- Ideogram: Ideogram 4.0 technical details (read 2026-10-04)
- Ideogram: licensing (read 2026-10-04)
- Sume docs: Image API
Related posts
More in Comparisons
- Fliki 180 credits is 60-90 minutes: what that costs on Sume
Fliki Standard lists 180 credits for about 60-90 minutes. Price 60, 75 and 90 minutes of avatar video on Sume by tier, and see which Fliki plan lists an API.
- FLUX.2 [klein] needs about 13 GB VRAM: local run or API?
Recraft cites about 13 GB of VRAM for FLUX.2 [klein] under Apache 2.0. A checklist for deciding between a local GPU and a hosted image API.
- FLUX.1 schnell $3 per 1,000 images: Replicate, fal and Sume's floor
FLUX.1 schnell is $3.00 per 1,000 images on Replicate and $0.003 per megapixel on fal. Sume's catalog has no schnell id; its lowest rows are $0.025 per image.
- Gemini 3.8 Live or TTS plus Fabric for a talking clip on Sume
Google shipped speech-to-speech and TTS models in September. A talking video on Sume is a TTS voice plus a Fabric clip. Why the two are different products.
Written by Sume