Kling Image 3.0 lists native 4K: Sume's image resolution tiers
Kling's page lists Kling Image 3.0 with native 4K. Sume's image API takes resolution 512, 1K, 2K or 4K per model; check the catalog before asking for 4K.

Kling AI's home page lists Kling Image 3.0 with native 4K output. We found no Kling image row in Sume's image docs we read, but its image API does accept a resolution of 512, 1K, 2K or 4K, subject to each model. The catalog carries 4K pricing rules for nano-banana-2.1 and nano-banana-pro. Always read a model's supported_parameters before you ask for 4K, because the API rejects a parameter that the selected model does not list.
What each page says
The Kling page lists Kling Image 3.0 with native 4K output. It gives no price or API detail in the text we read.
Sume's image docs list the tier values for resolution and add that size is tier shorthand only and rejects WxH; custom pixels go on image_size for models that support it. higgsfield/soul is the exception, using 720p or 1080p.
| Item | Kling Image 3.0 (Kling page) | Sume image API |
|---|---|---|
| 4K output | Native 4K | resolution 4K where the model lists it |
| Tier values | Not stated | 512, 1K, 2K, 4K (0.5K aliases 512) |
| Custom pixel size | Not stated | image_size on GPT, Seedream, Flux, Qwen, Recraft rows |
| Cost at 4K | Not stated | Docs note 4K scaling on nano-banana-2.1 and nano-banana-pro |
| Multiple images per request | Not stated | n, 1 to 10 plus a per-model cap |
Ask for 4K without a surprise
The cheapest habit is a catalog read. Call the images models endpoint, find your row, and confirm that resolution is in its supported parameters. Then price one image before you queue a hundred.
Steps:
- Request the Sume model list and filter to rows whose parameters include
resolution. - Submit one 4K image and read its cost.
- Scale the number up only after you like the output.
- Use
nfor series work, within the per-model cap.
curl -X POST https://api.sume.com/v1/images \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"nano-banana-pro","prompt":"A ceramic teapot on linen, soft window light","resolution":"4K"}'Why the tier name is not the whole story
A tier such as 4K is a label for the output size, not a promise about how the pixels were made. Some rows render natively at that size and some scale up, and Sume's docs describe the 4K price on two Nano Banana rows as scaled from a 1K list rate. When fine texture matters, such as fabric, small print or skin, zoom to 100 percent on a sample before you approve a batch.
Also keep the aspect ratio and the resolution separate in your head: the first sets the shape, the second sets the size tier, and a row can support one and not the other.
Budget a test first
Generate one image at the default tier and one at 4K from the same prompt, then compare the cost and the detail side by side. The difference tells you whether the extra tier earns its place in your workflow, and it takes two requests.
What Sume does not do
Sume does not list Kling Image 3.0 and does not promise native 4K on every image row: some are fixed at a lower tier, and a row that does not list resolution will reject the field. It also does not guarantee that a 4K tier is natively rendered rather than upscaled for each provider, so inspect the output when fine detail matters.
Sources
Related posts
More in Models
- LTX-2.5 license changed Aug 11, 2026: what differs from January
The August 11, 2026 LTX-2.x license adds a Non-Commercial carve-out, a current-version clause and provenance duties. What both texts say, side by side.
- LTX-2.5 quick start downloads 66 GiB: the five files and a disk plan
The LTX-2 README quick start pulls about 66 GiB for five LTX-2.5 files, including a 12B Gemma text encoder. What to budget, and what Sume lists instead.
- Mac or Windows team: which open video models need Linux and CUDA?
HunyuanVideo and HunyuanVideo-1.5 list Linux; LTX-2.5's fastest decoder is Linux plus CUDA; H3 has a Mac route. What each page says and a hosted option.
- MAI-Transcribe-2: 60 languages, 17 more than 1.5, and the Sume hint
Microsoft lists 60 languages for MAI-Transcribe-2, 43 for MAI-Transcribe-1.5, so 17 new ones. Sume STT takes a language_code hint or auto-detects.
Written by Sume