Cost of 200 five-second clips a month: hosted Sume ids vs a rented GPU
At Sume prices, 200 five-second clips cost $75 to $200 a month depending on id and resolution. The same sum for a rented GPU, with the break-even hours.
At 200 five-second clips a month, Sume's hosted video ids cost $125.00 on wan-3.0 at 720p, $75.00 on minimax-h3 at 768p, and between $100.00 and $200.00 on minimax-h3-max at 768p or 1080p. A rented GPU wins only if the hours you pay for are fewer than the break-even in the second table. Prices are Sume's per-second rate, provider list times 1.25, as of 2026-10-08.
Monthly cost by id and volume
Per-clip price = 5 seconds x the Sume per-second rate. Monthly price = clips x per-clip price.
| Sume model id | Resolution | Per second | Per 5 s clip | 50 clips | 200 clips | 1,000 clips |
|---|---|---|---|---|---|---|
| wan-3.0 | 720p | $0.125 | $0.625 | $31.25 | $125.00 | $625.00 |
| minimax-h3 | 768p | $0.075 | $0.375 | $18.75 | $75.00 | $375.00 |
| minimax-h3-max | 768p | $0.10 | $0.50 | $25.00 | $100.00 | $500.00 |
| minimax-h3-max | 1080p | $0.20 | $1.00 | $50.00 | $200.00 | $1000.00 |
The rented-GPU side
Open-weights serving has no per-clip price, only GPU time. To compare, take a measured seconds-per-clip on your hardware, multiply by GPUs, and divide by 3,600 for GPU-hours. MiniMax's page gives one reference: about 87 seconds on 4 B300 GPUs for an 8.7 second H3 clip, which is 348 GPU-seconds. The hourly rate below is an assumption for illustration, not a vendor price.
| Hosted monthly spend (200 clips) | Spend / $4.00 per GPU-hour | GPU-hours you can rent before hosted is cheaper |
|---|---|---|
| $125.00 | 125.0 / 4.00 | 31.2 |
| $75.00 | 75.0 / 4.00 | 18.8 |
| $100.00 | 100.0 / 4.00 | 25.0 |
How the hosted price is built
Sume's video docs say that at submit Sume reserves the provider list price times 1.25 for each model, including minimax-h3-max. The poll response reports usage.cost as the Sume billable amount. So a 5 second wan-3.0 clip at 720p reserves 5 x $0.10 x 1.25 = $0.625, and the table scales that by clip count.
Duration limits differ: wan-3.0 accepts 2 to 30 seconds, minimax-h3 and minimax-h3-max accept 5 to 15.
Reading the break-even
At the assumed rate, the 200-clip Wan 3.0 spend equals 31.2 GPU-hours, which is 112,500 GPU-seconds. Divided by 200 clips, that is 562 GPU-seconds per clip. If your measured number is higher than that, hosted is cheaper, before counting idle time and engineering. The H3 weights license excludes the US, EU, UK and South Korea, so that route is not open to every team.
What to confirm before you rely on the numbers
Sume reserves the Sume price at submit and the poll response reports the billable amount, so the table is a forecast you can check against real jobs.
- Run GET /v1/videos/models for the current resolutions and durations; limits differ by model.
- Confirm the list price has not moved: the repo table behind this post was read on 2026-10-08.
- Add upscales: Sume bills 2K or 4K upscales on
minimax-h3if the request includes them, and this table does not.
Sources
Related posts
More in Pricing
- One minute of AI video: Runway Gen-4.5, Luma Ray3.2, Wan 3.0, Seedance
One minute of AI video costs $3.60 to $85.29 depending on the model: Runway $7.20, Luma Ray3.2 $3.60 to $14.40, Wan 3.0 $7.50 on Sume. Arithmetic inside.
- One minute and one hour of AI narration: MAI-Voice vs Sume TTS cost
At 900 characters a minute, narration costs about 2 cents on MAI-Voice-2.1, 1.4 cents on Flash, 5 cents on Sume (job rounding). One hour: $1.19, $0.81, $2.58.
- What a default sume/auto video costs: 8 seconds at 720p is $1.00
Sume Auto video defaults to 720p and 8 s, and the default model is Gemini Omni Flash 1.1: $0.10 list per second x 8 x 1.25 = $1.00. 100 calls cost $100.
- Do Omni references cost extra? Ten images bill like none
On Sume, Gemini Omni Flash 1.1 bills per output second by resolution. A clip with 10 reference images and 3 clips costs the same as one with none.
Written by Sume