10 s of Omni Flash at 720p: $1.25 on Sume vs ~$1.01 at Google list
Google lists Gemini Omni Flash video output at $17.50 per 1M tokens, about $0.10 a second at 720p. Sume's 720p rate is that times 1.25. The 10-second sum.
Ten seconds of Gemini Omni Flash 1.1 at 720p costs $1.25 on Sume. Google's Gemini API pricing page lists the model's video output at $17.50 per 1M tokens with 5,792 tokens per second of 720p video, about $0.10 per second, and no free tier (read 2026-10-08). That works out to $1.0136 for ten seconds at list.
The gap is the Sume rule: provider list times 1.25. $0.125 per second is $0.10 times 1.25.
The arithmetic
Only 720p has a Google figure on the page I read, so the comparison stops there.
| Step | Value |
|---|---|
| Google video output rate | $17.50 per 1M tokens |
| Tokens per second of 720p | 5,792 |
| Google per second | $0.1014 |
| Google, 10 s | $1.0136 |
| Sume per second (list $0.10 x 1.25) | $0.125 |
| Sume, 10 s | $1.25 |
| Difference per 10 s clip | $0.2364 |
What the difference buys
On Sume the model sits behind the same /v1/videos request, balance, Idempotency-Key and job polling as every other video model, so switching to Seedance or Wan is a model-id change. Billing is a workspace USD balance reserved at submit.
If you only ever call this one model and already hold a Google billing account, the list price is lower per second. If you want one key and one bill across the catalog, the premium is the price of that.
At volume
At 1,000 ten-second 720p clips the list gap is $236.40. Run your own count before you decide; prices on both sides are subject to change.
A shorter clip, same ratio
At 5 seconds and 720p the same arithmetic gives $0.5068 at Google list and $0.625 on Sume, a $0.1182 gap. The ratio does not change with length, so the premium is always 25% of the provider rate.
Gemini Omni Flash 1.1 on Sume accepts 3 to 10 seconds at 360p, 720p, 1080p and 4K in 16:9 or 9:16, with native synced audio (Video generation). Because only the 720p figure was on the Google page I read, every other resolution comparison is left out on purpose.
Sources
Related posts
More in Comparisons
- 12 reference photos in one edit: only GPT Image 2.5 takes them
Nano Banana 2.1 takes 10 references on Sume, Ideogram 4.5 takes 5, GPT Image 2.5 takes 16. For a 12-photo edit use GPT Image 2.5, or split the job in two.
- 12 s at 720p: Seedance 2.5 $6.94, 2.0 $4.54, Fast $3.63, Mini $2.27
Four Seedance rows priced for the same 12-second 720p clip on Sume, with the list rate per 1000 tokens and when to pick each.
- 12 s clip: sume/auto caps at 10 s; pin Wan $1.50 or Seedance 2.5 $6.94
sume/auto takes 3 to 10 s, so a 12 s clip needs a pinned model. 12 s at 720p on Sume: H3 $0.90, Wan $1.50, Kling $1.68, Mini $2.27, Seedance 2.5 $6.94.
- 12-second motion transfer: Genjutsu ~$4.77 vs Kling motion $1.89
Moving motion onto a character for 12 seconds: Kling 3.0 Motion Control is $1.89 on Sume; Genjutsu, where enabled, is about $4.77 at 480p. Inputs differ.
Written by Sume