MiniMax H3 defaults to 2K at MiniMax; on Sume the default is 768p
MiniMax's H3 page says 2K is the default. Sume's minimax-h3 runs native 480p or 768p with a 768p default; 2K and 4K are billed upscales.

MiniMax's own H3 announcement, dated July 31, 2026, says the model produces clips of up to 15 seconds with 2K as the default resolution, 768p also supported and native stereo sound. On Sume, minimax-h3 works differently by default: its public resolutions are native 480p and 768p, the product default is 768p, and Sume bills 2K or 4K upscales only if your request includes them. If you port a MiniMax call to Sume and leave resolution out, you get 768p, not 2K.
What each page says
The MiniMax page makes relative price claims against other models, which we do not repeat. It also says open weights are planned.
Sume's video guide says minimax-h3 accepts 5 to 15 seconds at native 480p or 768p, that 768p is first-class and not 720p, and that Sume bills 2K or 4K upscales if your request includes them. H3 Max is a separate Sume row for 480p, 768p and 1080p.
| Item | MiniMax H3 page | Sume minimax-h3 |
|---|---|---|
| Default resolution | 2K | 768p |
| Other resolutions named | 768p | 480p; 2K and 4K upscales if requested |
| Clip length | Up to 15 seconds | 5 to 15 seconds |
| Audio | Native stereo | Native stereo, no toggle |
| Price at 768p | Relative claims only | $0.075 a second (list $0.06 x 1.25) |
| Price at 480p | Not stated | $0.0625 a second |
Pin the resolution explicitly
Never depend on a default when a model's default differs between providers. Set resolution in every request so the price and the output size are what you planned. A 10-second clip at 768p is $0.75 on Sume; the same clip at 480p is $0.63 after rounding up to the cent.
Steps:
- Decide the delivery size first: 768p is enough for most vertical social clips.
- Send
resolutionin the body, as768p, never as720p. - Check the 2K and 4K rates in the catalog before you request them; they are priced higher.
- Keep
minimax-h3-maxin mind if you need 1080p.
{
"model": "minimax-h3",
"prompt": "A paper boat crosses a rain puddle, macro, shallow focus",
"resolution": "768p",
"duration": 10,
"aspect_ratio": "9:16"
}When a default matters
Defaults hide in two places: cost and size. A pipeline that expects 2K frames will break a crop or compositing step when it receives 768p files, and a budget built on a 768p rate will be wrong if someone later asks for an upscale. Write the resolution into your request template, and record which tier each batch used.
If you need native 1080p from the H3 family on Sume, the catalog row for minimax-h3-max lists 480p, 768p and 1080p, at $0.0625, $0.10 and $0.20 per second. Its 1080p is described in the docs as a latent refinement from 768p, so judge a sample before committing a batch.
Checking your own call
Fetch the row with GET /v1/video-router/models/minimax-h3 and read its capabilities before you hard-code anything. The row shows the resolutions and durations the API will accept, and an unsupported value returns an error rather than a silent downgrade.
What Sume does not do
Sume does not default to 2K for H3 and does not endorse the cost comparisons MiniMax makes against other models. Through Sume you pay the per-second rate above.
Sources
Related posts
More in Models
- MiniMax H3 on a 12 GB RTX 3060: 480p only, 42 GB, and the price
MiniMax says 12 GB of VRAM runs 480p H3 clips with audio from a pruned int8 checkpoint of about 42 GB. What that buys, and a 480p clip on Sume costs.
- Mistral Large 4 open weights: Oct 31 ETA and what to do now
Mistral's own Hugging Face page lists Mistral Large 4 as upcoming with an October 31 ETA. What is known, what is not, and why Sume's agent run is unaffected.
- Moroccan Arabic speech to text: open-source Voxtral model vs Sume STT
A Moroccan open-source stack pairs language ID with Voxtral ASR for dialect code-switching. What Sume STT offers for Arabic audio, and what it lacks.
- Nano Banana 2.1 or Pro: 10 vs 19 cents, and four ratios Pro lacks
On Sume, Nano Banana 2.1 bills 10 cents and Pro bills 19. Both take edits and four resolution tiers, but Pro does not list 4:1, 1:4, 8:1 or 1:8.
Written by Sume