FLUX 3 API price per image: 768sq to 4k, and pricing on Sume
BFL lists FLUX 3 Image at $0.041 (768sq), $0.048 (1k), $0.100 (2k) and $0.607 (4k) per image. Sume prices come from GET /v1/images/models.

BFL's release notes, read 2026-10-01, price FLUX 3 Image per image by resolution: $0.041 at 768sq, $0.048 at 1k, $0.100 at 2k and $0.607 at 4k. Sume does not list FLUX 3 Image in its image catalog (the FLUX entries in code are flux-2-pro and flux-2-flex), so for Sume you read the price of the model you will actually call from GET /v1/images/models.
BFL numbers are from its release notes; Sume facts from the Image API, read 2026-10-01.
What does FLUX 3 Image cost at each size?
The ratio between the ends is large. Using only the table below, a 4k image costs $0.607 against $0.048 at 1k, which is about 12.6 times as much, and 100 images come to $4.80 at 1k or $60.70 at 4k.
| `resolution` | Price per image | 100 images |
|---|---|---|
768sq | $0.041 | $4.10 |
1k | $0.048 | $4.80 |
2k | $0.100 | $10.00 |
4k | $0.607 | $60.70 |
How do I price an image on Sume?
The Image API docs say you can explore available models and pricing with GET /v1/images/models, and that endpoint pricing lines are the amount charged to your wallet, with Sume's margin already applied, so cost_usd times n is what you pay. Sume's resolution tiers are 512, 1K, 2K and 4K. Read the line for your model and tier rather than assuming one price per model.
Do not mix the two tables: BFL's numbers are its list prices for its endpoint, and a Sume line is the wallet charge for a different model. The Sume side of the markup is explained in image API prices at list times 1.25.
Do failed generations cost anything on Sume?
The docs say image generation billing is all-or-nothing: a generation is either completed and billed in full, or it fails and is not billed. Failed or canceled generations are not billed, and a client that disconnects early is billed as a failed generation, which is not at all.
What should I do before committing to a size?
Run the same prompt at the lowest and highest tier you might ship, compare the outputs, and multiply by your volume using the price you read for that model. A cheaper preview size is a reasonable first pass, as the BFL notes themselves describe 768sq as the fast preview size. For the generate-versus-upscale trade at 4K, see 4K image: generate or upscale.
Sources
Related posts
More in Pricing
- FLUX API: 1 credit = $0.01 at BFL, dollars in Sume usage.cost
BFL prices FLUX in credits at $0.01 each. Sume reports the billed USD amount in usage.cost and exposes GET /v1/balance and GET /v1/usage, so no conversion.
- Gemini Omni Flash edit billed per input second vs Sume reserve
Runway bills Gemini Omni Flash video-to-video at 11 credits per input second, capped at 10 seconds. Sume reserves an edit from an 8s hint, not a probe.
- gpt-image-2.5 batch pricing: $4 image input, $15 output per 1M
OpenAI's batch rate for gpt-image-2.5 is $4.00 image input and $15.00 output per 1M tokens. Sume's docs describe a per-image reserve, not a batch tier.
- gpt-realtime-translate pricing: $0.034 per minute, and STT plus TTS
OpenAI lists gpt-realtime-translate at $0.034 per minute. Sume has no live translate line; STT 1.0 and TTS 1.0 are separate rates you add up yourself.
Written by Sume