Imagen 4 Fast 3 cents vs Ultra 8 cents: text-only, five ratios on Sume
Imagen 4 Fast bills 3 cents and Ultra 8 cents on Sume. Both are text-to-image only with five ratios; Ultra adds 1K and 2K tiers. Cost of 100 to 10,000 images.

Imagen 4 Fast bills 3 cents per image on Sume and Imagen 4 Ultra bills 8 cents. Both are text-to-image only, so image_urls and input_references are rejected, and both list the same five ratios: 1:1, 16:9, 9:16, 4:3 and 3:4. Ultra adds a resolution choice of 1K or 2K.
Cost at four volumes
Catalog list prices are $0.02 (Fast) and $0.06 (Ultra), times 1.25 gives $0.025 and $0.075, rounded up to 3 and 8 cents. Rounding matters: Fast's real 2.5 cents becomes 3 cents billed, so 10,000 images cost $300, not $250.
| Model | Billed per image | 100 | 500 | 1,000 | 10,000 |
|---|---|---|---|---|---|
| Imagen 4 Fast | 3 cents | $3 | $15 | $30 | $300 |
| Imagen 4 Ultra | 8 cents | $8 | $40 | $80 | $800 |
Where these rows fit
The limits are plain. There is no 4:5 and no edit mode, so a product shot from a reference photo, an Instagram portrait or a text revision cannot be done on Imagen. What Imagen does give you is a cheap, simple text-to-image call with common landscape and portrait shapes, plus a 2K option on Ultra.
If you need 4:5 at a similar price, Qwen Image and Flux 2 Pro list it at 3 and 4 cents. If you need to edit, move to one of the 14 edit-capable rows.
- Fast for drafts: 100 images are $3.
- Ultra for finals at 1K or 2K: 100 images are $8.
- Anything with a reference photo: pick a different row.
A short check
Before you build on Imagen, send one request with your ratio. A ratio outside the five returns 400 unsupported_parameter, which is cheaper to find out now than after a batch of 500.
A drafting loop
Use Fast for 8 prompt ideas at 4 images each: 32 images and 96 cents. Pick the two prompts that work, then run those on Ultra at 1K, 4 images each, for 64 cents. The whole loop costs $1.60 and produces 8 candidate finals. Add 2K on Ultra only for the one you will print or crop.
If your pipeline sometimes carries a reference photo, put an edit-capable row behind a router in your own code, and send references only to rows whose input_references maximum is above zero.
Sources
Related posts
More in Models
- Inworld buys Ultravox, voices move to TTS-2: pin voice and model IDs
Inworld's Ultravox deal moves built-in voices to Realtime TTS-2 at no extra cost. Why to pin voice and model ids wherever you generate speech.
- Is Hy Image 3.5 Preview on Sume? Check the models list in code
Tencent's Hy Image 3.5 Preview is not in the Sume rows used here. How to confirm against GET /v1/images/models with a 12-line script instead of guessing an id.
- Is MAI-Voice-2.1 on Sume? What Sume's audio docs list instead
Sume does not list MAI-Voice-2.1 or MAI-Transcribe-2. Here is what its docs do list for speech, transcripts, music and captions, with prices.
- Kling 3 audio on vs off for 100 fifteen-second SKU ads: $105 gap
Kling 3 Pro on Sume is $0.14 per second with audio off and $0.21 on. At 15 s across 100 SKUs that is $210 vs $315, a $105 difference.
Written by Sume