MiniMax H3 768p 'half the price' claim checked against Sume rows

On Sume's price list the MiniMax claim does not hold against every model. MiniMax says its H3 at 768p costs less than half of mainstream models' 720p (read 2026-10-04). Sume bills H3 at 768p at $0.075 per second. Wan 3.0 and Gemini Omni Flash 1.1 at 720p are $0.125, and Kling 3 is $0.14, so H3 is 40 to 46 percent cheaper than those three, not half.

Against Grok Imagine Video 1.5, H3 is more expensive: $0.0125 per second at 480p or 720p.

What is each Sume row per second?

Billed price is the provider list times 1.25. The table lists the closest tier to 720p for every per-second model.

Why can the claim still be true for MiniMax?

MiniMax compares against an unnamed group of mainstream 720p models on its own page, and Sume's catalog may not match that group. The comparison that matters on Sume is Sume's own rows, as above, and that is the one to use when you budget.

H3 also differs in what you get: native stereo audio and a 5 to 15 second range. On Sume the 1080p tier belongs to minimax-h3-max, not the base H3 row.

Which row should you pick?

If audio and a 5-second floor suit the shot, H3 at 768p is the cheapest per-second row that is not an image-only model. If you need image-to-video only and the cheapest rate, Grok is lower. If you need references or 30 seconds, move to Wan or Seedance.

• 5-second H3 clip at 768p: $0.38 billed.
• 5-second H3 Max clip at 768p: $0.50 billed; at 1080p $1.00.
• 5-second Wan 3.0 clip at 720p: $0.63 billed.

How do you check the rate yourself?

Read the model list and compute from pricing_skus, or run one job and read usage.cost.

import os, requests

r = requests.get(
    "https://api.sume.com/v1/videos/models",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
)
for m in r.json()["data"]:
    if m["id"].startswith("minimax"):
        print(m["id"], m["supported_durations"][:2], m["pricing_skus"])