MiniMax H3 Max for 14 seconds: $0.88 at 480p, $2.80 at 1080p on Sume

A 14-second minimax-h3-max clip on Sume costs $0.88 at 480p, $1.40 at 768p and $2.80 at 1080p. The row accepts 5 to 15 seconds and always produces native stereo audio. Its 1080p output is a latent refinement of native 768p, not a native 1080p render, which is why the 768p tier is the one to evaluate for quality first.

Price table

Sume bills the provider list price times 1.25 for this row, as for every video model. The 768p and 1080p tiers double in price from one to the next.

Modes on this row

H3 Max does text-to-video, first and last frame image-to-video, and reference-to-video. The input references include image, video and audio types. The catalog constraints say up to 9 reference images, up to 3 reference videos and up to 3 reference audio clips (2 to 15 seconds each, 15 seconds combined), at most 12 items in total, and audio cannot be the only reference.

Reference-to-video is billed as output seconds at list times 1.25; the catalog notes that the provider's reference-token overage is not reserved.

• Aspect ratios: adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16.
• There is no generate_audio toggle; stereo audio is always produced.
• Duration: 5 to 15 seconds, whole seconds.

H3 or H3 Max

The plain minimax-h3 row costs $0.075 per second at 768p, against $0.10 for H3 Max, and stops at 768p. For 14 seconds that is $1.05 against $1.40. Pay the 35 cents difference when you want the faster 768p variant or the 1080p tier; otherwise H3 is the cheaper way to get the same 14 seconds.

Draft strategy

Duration is the multiplier you control. A 5-second 768p draft costs 5 x $0.10 = $0.50, so you can test the prompt for about a fifth of the cost of the 14-second 768p final ($1.40) before committing. Because H3 Max accepts 5 to 15 seconds, the draft and the final can use the same settings except duration.

Submit with an Idempotency-Key per job. If the poll times out on your side, resubmitting with the same key returns the original job rather than starting a second paid generation.