Kling 3.0 4K in Sume's Videos panel: what actually gets submitted
The panel lists 4K for Kling 3.0 but its live submit maps 4K down to 1080p. Where 4K does exist on Sume's API, and how to check the model before you pay.

In Sume's Agents Videos panel, Kling 3.0 lists 4K as a resolution, but the live create path maps a 4K choice down to 1080p before it submits. The code comment gives the reason: the v1 path rejects or reserves 4K. If you need real 4K, check the catalog entry for the model first; Sume's docs list 4K for Gemini Omni Flash 1.1.
This is a gap between a panel control and what is sent, and the safest way to see the truth is the finished job, not the chip.
What does the panel do with the 4K chip?
The panel's Kling 3.0 entry offers 720p, 1080p and 4K. When it builds the request, it keeps your resolution except that 4K becomes 1080p for every model other than the two MiniMax entries. Auto, Wan 3.0 and Grok Imagine never offered 4K in the panel, so Kling 3.0 is the only entry where the 4K option can be mapped down.
The panel's mock credit estimate (used when the live path is not available) adds credits for 4K, while the live estimate prices the 1080p that is actually submitted. Neither is a bill: read the real amount from the job's usage.cost.
| Model | 4K chip in panel | Submitted resolution when 4K is chosen |
|---|---|---|
| Auto | Not listed | Not applicable |
| Kling 3.0 | Listed | 1080p |
| Wan 3.0 | Not listed | Not applicable |
| MiniMax H3 | Not listed | Not applicable |
| Grok Imagine | Not listed | Not applicable |
Does Kling itself offer 4K?
Yes: Kling's 4.0 versus 3.0 comparison says both versions support 720p, 1080p and 4K, with 4.0 adding 10-bit HDR at 1080p and 4K. So the vendor offers 4K on 3.0; the narrowing is in Sume's panel path, which this post is about. Check the Kling 3 4K API post for the API angle before deciding.
Where does 4K exist on Sume?
The video docs list Gemini Omni Flash 1.1 (gemini-omni-flash-1.1) at 3 to 10 seconds across 360p, 720p, 1080p and 4K in 16:9 or 9:16, with native synced audio. The legacy Video 1.0 URL rejects 4k and points you to POST /v1/videos instead, and MiniMax H3 prices 2K and 4K upscales when you request them.
List the catalog and read the resolutions yourself:
curl "https://api.sume.com/v1/videos/models" \
-H "Authorization: Bearer $SUME_API_KEY"
# look for "4K" in supported_resolutions on the model you plan to useHow do I avoid paying for 4K I did not get?
Do not trust the chip. Submit one short clip, open the finished job, and check the output size and the cost on the job before queuing a batch. If the result is 1080p, either accept it or move to a model whose catalog entry lists 4K.
Gemini Omni Flash 1.1 lists 3 to 10 seconds in Sume's docs, and a separate post covers an 8-second limit at 4K, so confirm the duration allowed at 4K before you plan a longer clip.
What should I check before a batch?
For anything priced or delivered as 4K, the check is the output file itself: its pixel size, not the chip you selected.
Submit one short clip first, open the finished job, and compare what you asked for with what came back. Use an Idempotency-Key on each attempt, because a replay with the same key returns the original job instead of creating and billing a second one. Only then queue the rest.
Sume reserves provider list times 1.25 when a job is submitted, and the poll response's usage.cost is the billable amount. Treat that field, not a panel estimate, as the number to budget with.
What does Sume not do?
Sume does not promise 4K from Kling 3.0 through the panel, and Sume's docs do not list Kling 4.0. Kling says Kling 4.0 is planned for wider rollout in October 2026; until it is in Sume's catalog, plan on what the catalog returns today.
Sources
Related posts
More in Models
- Kling motion control keep_original_sound: get a silent clip
keep_original_sound on Sume's Kling 3.0 Motion Control defaults to true, so the driving video's audio rides along. Send false for a silent clip.
- LTX-2.3 LoRAs on LTX-2.5: most run unchanged; Sume has no LoRA field
Lightricks says the large majority of LTX-2.3 LoRAs and IC-LoRAs run on LTX-2.5 unchanged, with a few exceptions to test. Sume has no LoRA field.
- LTX-2.5 distilled or dev transformer: which file to download
Download the distilled bf16 transformer to generate (fixed 8-step schedule, CFG 1) and the dev one to train. int8 files are ComfyUI-only. Sume lists no LTX id.
- LTX-2.5 duration predictor: optional, and Sume sets seconds
LTX-2.5 has an optional duration predictor that sets the frame count from the prompt. Sume has no such mode: you send duration in whole seconds.
Written by Sume