MiniMax H3 or H3 Max after Sora: native 768p vs latent 1080p prices
On Sume, H3 renders native 480p or 768p and bills 2K/4K upscales; H3 Max adds 1080p as a latent refinement of 768p. 10-second prices side by side.

Pick minimax-h3 when 768p native is enough and you want the lowest price: it is about $0.075 a second on Sume at 768p. Pick minimax-h3-max when you need 1080p: it is the faster 768p variant and produces 1080p as a latent refinement of the native 768p, at about $0.20 a second. Both take 5 to 15 seconds. Neither is a quality ranking; they are two products with different resolution paths.
OpenAI retired the Sora Videos API on 2026-09-24 (Pondero). The H3 pair is one of several ways to replace it, and the question most migrations hit is which of the two to point at.
What the docs say
minimax-h3 accepts 5 to 15 seconds at native 480p and 768p, and the docs stress that 768p is first-class, not 720p. Sume bills 2K and 4K upscales if the request includes them. minimax-h3-max is the faster 768p variant: it does text-to-video, first/last-frame image-to-video and reference-to-video at 480p, 768p and 1080p for 5 to 15 seconds, with native stereo audio, and reports image, video and audio input references.
Prices for the same 10-second clip
Prices below use the list per second in Sume's pricing code, times the documented 1.25 margin.
| Model | Resolution | List per s | Sume per s | 10 s clip |
|---|---|---|---|---|
| minimax-h3 | 480p | $0.05 | $0.0625 | $0.63 |
| minimax-h3 | 768p | $0.06 | $0.075 | $0.75 |
| minimax-h3-max | 480p | $0.05 | $0.0625 | $0.63 |
| minimax-h3-max | 768p | $0.08 | $0.10 | $1.00 |
| minimax-h3-max | 1080p | $0.16 | $0.20 | $2.00 |
How to choose
At 768p the same ten seconds cost $0.75 on H3 and $1.00 on H3 Max, a 33 percent premium. The reason to pay it is the extra capabilities rather than the pixels: H3 Max is the one that the docs list with 1080p, and with first/last-frame image-to-video. If a 768p vertical clip is your whole need, H3 is the cheaper route.
A caution on 1080p: because it is a latent refinement of native 768p, it is not a native 1080p render. If your Sora outputs were native 1080p, compare the two on a real prompt before you move a whole batch.
Extras that change the bill
H3 has a reference-image rule in Sume's pricing code: five reference images are included and each additional one adds a list charge. The docs do not restate it, so confirm in the catalog's pricing_skus before you send a big reference set.
The default resolution on H3 is 768p in Sume's code, not the 2K upscale some providers default to. That keeps an unset resolution from silently costing more.
A short test plan
Take three prompts that represent your Sora workload: a product shot, a person, a moving camera. Render each at 768p on both models for 10 seconds. That is $0.75 times 3 plus $1.00 times 3, which is $5.25 in total at the table prices. Then render the winner at 1080p on H3 Max for another $2.00 per prompt if you need the larger frame. Read the clips side by side before you commit a batch to either route.
Remember audio: H3 Max has native stereo audio, so a pipeline that needs silence should plan for a post-step.
Sources
Related posts
More in Models
- Minimum clip length by Sume video model: 2, 3, 4 or 5 seconds
Wan 3.0 starts at 2 s, Gemini Omni Flash 1.1 at 3 s, Seedance and Genjutsu at 4 s, MiniMax H3 and Recast at 5 s. Full min and max table for a Sora port.
- Music Router image_url: public HTTPS only, and null to clear it
Sume music image_url must be a public HTTPS image, and null clears it on a reused request object. When a mood still helps and when the prompt does more.
- Nano Banana 2 is retired on Sume: same price on Nano Banana 2.1?
Sume retired google/nano-banana-2 and runs old requests as Nano Banana 2.1 at the same price card: $0.10 at 1K billed. Tiers, ids and what the job stores.
- Native 4K or 2K image edits: pixel ceilings by model
How many pixels each model reaches: Flux 3 Image near 16.8 MP, GPT Image 2.5 capped at 8,294,400 on Sume, Ideogram 4.5 at 2K. What Sume serves.
Written by Sume