Edit with input_references: 5 of 19 Sume image models are text-only
Five Sume image models cannot take input_references: Soul, Imagen 4 Fast, Imagen 4 Ultra, Recraft V4 and Qwen Image Max. Here are the 14 that can, with prices.
Five of the 19 Sume image models are text-to-image only and reject input_references: Higgsfield Soul, Imagen 4 Fast, Imagen 4 Ultra, Recraft V4 and Qwen Image Max. The other 14 accept references for edits, from Grok Imagine and Qwen Image at 3 cents up to ChatGPT Image 2 at 27 cents.
The five that cannot edit
In the catalog these rows have input_references set to min 0, max 0. Sume describes it plainly: if the descriptor is {"min": 0, "max": 0}, the model is text-to-image only and rejects references. Plan for this before you build an edit pipeline, because Qwen Image Max and Qwen Image are different rows with different abilities.
| Model | Billed per image | Aspect ratios listed | Note |
|---|---|---|---|
| Soul (Higgsfield) | 1 cents | 7 | text-to-image only; n is 1 or 4 |
| Imagen 4 Fast | 3 cents | 5 | text-to-image only; image_urls not accepted |
| Recraft V4 | 5 cents | 13 | text-to-image only; image_urls not accepted |
| Imagen 4 Ultra | 8 cents | 5 | text-to-image only; image_urls not accepted |
| Qwen Image Max | 10 cents | 13 | text-to-image only; image_urls not accepted |
The 14 that can
Reference images are public HTTPS URLs. Sume rejects localhost, private-network and non-HTTPS URLs before it submits the job. On edit calls, aspect_ratio: auto matches the reference, and omitting the field is not the same thing.
| Model | Billed per image | 100 edits | 1,000 edits |
|---|---|---|---|
| Grok Imagine | 3 cents | $3 | $30 |
| Qwen Image | 3 cents | $3 | $30 |
| Flux 2 Pro | 4 cents | $4 | $40 |
| Seedream 4.0 | 4 cents | $4 | $40 |
| Seedream 4.5 | 5 cents | $5 | $50 |
| Seedream 5.0 Lite | 5 cents | $5 | $50 |
| Flux 2 Flex | 7 cents | $7 | $70 |
| ChatGPT Image 2.5 | 7 cents | $7 | $70 |
| ChatGPT Image 2.5 Sunburst | 7 cents | $7 | $70 |
| Ideogram V3 | 8 cents | $8 | $80 |
| Ideogram 4.5 | 8 cents | $8 | $80 |
| Nano Banana 2.1 | 10 cents | $10 | $100 |
| Nano Banana Pro | 19 cents | $19 | $190 |
| ChatGPT Image 2 | 27 cents | $27 | $270 |
How to choose
If you need many references, use GPT Image 2.5 (up to 16). If you need a text edit on an existing graphic, Ideogram 4.5 edits the first reference and takes up to 4 more. For a cheap colorway or background swap, Grok Imagine and Qwen Image at 3 cents are the least expensive entries. Read the live descriptor with GET /v1/images/models before you hard-code a limit.
Handling the rejection
A text-only row that receives input_references answers with a 400, which is why it pays to read the catalog before a batch begins. The check is cheap: list the models once, build a map of model id to the input_references maximum, and route any job with references only to rows where the max is above zero. A fallback from a text-only row to an edit-capable row needs a prompt rewrite as well, because an edit prompt describes a change, while a text-to-image prompt describes the whole scene.
Sources
Related posts
More in Models
- EU, UK or South Korea team: which open video weights you may use
HunyuanVideo and MiniMax H3 licenses exclude the EU, UK and South Korea. Wan 2.2 is Apache 2.0 and LTX-2.5 is worldwide with a revenue line.
- Five requests Omni Flash cannot take, and what to send instead
Gemini Omni Flash 1.1 on Sume refuses clips over 10 seconds, silent clips, audio references, 1:1 and 2K or 480p output. What to send instead for each.
- Flux 2 Pro 4 cents or Flex 7: 1,000 images is $40 or $70 on Sume
Flux 2 Pro bills 4 cents and Flux 2 Flex 7 cents on Sume. Both list 13 ratios, including 21:9, and both edit. Cost of 100 to 10,000 images.
- FLUX 3 Image is on BFL's page: which FLUX id does Sume call?
BFL lists FLUX 3 Image, FLUX.2 Max and Klein. Sume's image router calls FLUX.2 pro and flex. Ids, prices at $0.0375 and $0.0625, ratios and edit limits.
Written by Sume