FLUX.2 Pro: fal bills per megapixel, Sume lists one flat $0.0375
fal prices FLUX.2 Pro at $0.03 for the first megapixel plus $0.015 per extra (read 2026-10-10). Sume's row is a flat $0.0375 with no resolution field.
FLUX.2 Pro costs a flat $0.0375 per image on Sume, with no resolution field to set, while fal's model card read on 2026-10-10 prices it at $0.03 for the first megapixel of output plus $0.015 for each extra megapixel. A budget built from the fal formula grows with image size; a budget built from the Sume line does not.
The $0.0375 is the $0.03 first-megapixel rate times the 1.25 margin in the pricing code. This post explains what that means for planning, and what it does not tell you, which is the pixel size Sume returns.
The two pricing shapes
On the fal page the formula is $0.03 for the first megapixel plus $0.015 per extra megapixel, with a minimum charge so that a 512x512 image rounds up to $0.03. If you assume whole megapixels with no other rounding, a 1 MP image is $0.03, 2 MP is $0.045, 3 MP is $0.06 and 4 MP is $0.075. The page does not state a rounding rule for fractional megapixels, so treat those as a pattern rather than exact invoices.
Sume's Image API docs say endpoint pricing lines already include the margin and you pay cost_usd x n. For black-forest-labs/flux.2-pro the catalog line is $0.0375 and for FLUX.2 Flex it is $0.0625.
| Output size | fal (whole-MP assumption) | Sume billed |
|---|---|---|
| 1 MP | $0.030 | $0.0375 |
| 2 MP | $0.045 | $0.0375 |
| 4 MP | $0.075 | $0.0375 |
What Sume does not let you set
The Sume catalog descriptors for FLUX.2 Pro list aspect_ratio (13 values), n from 1 to 4, input_references up to 10 and output_format, and no resolution. The docs are explicit that a request which sets a parameter the model does not list gets 400 unsupported_parameter. So you cannot ask Sume for a 4 MP FLUX.2 Pro image the way you can ask for 4K on Nano Banana 2.1.
I cannot tell you the pixel size a Sume FLUX.2 Pro image comes back at, because the docs do not state one and I did not generate images for this post. Read the dimensions of one returned file before you plan a print or a crop.
- Do not copy the fal per-megapixel formula into a Sume cost model.
- Do not send
resolutionto FLUX.2 Pro; it is not in the descriptor. - Use
aspect_ratioto choose the shape, and measure one returned file for size.
When the flat price helps
A flat price is easy to forecast. For 2,000 stills, Sume's FLUX.2 Pro line gives 2,000 x $0.0375 = $75.00 whatever shape you choose. The fal formula gives $60.00 at 1 MP and $150.00 at 4 MP under the whole-megapixel assumption, so the shape of your outputs matters there.
If you need guaranteed large output from a per-megapixel model, a row with a documented resolution field is the safer pick on Sume. Nano Banana 2.1 lists 512, 1K, 2K and 4K, and Ideogram 4.5 lists 1K and 2K.
A practical way to compare the two shapes
If your jobs vary in size, the cleanest test is to pick one prompt and run it on FLUX.2 Pro through each route, then compare the billed amount with the size of the file you got. On Sume that billed amount is the usage.cost field of the response, which the docs define as the USD amount billed to your wallet, and token counts there are always 0 because image rows are metered per image.
Keep the comparison honest by recording the date. The fal card I read is a live page, and the Sume line comes from a catalog that changes when the pricing code is edited. A comparison from 2026-10-10 is a snapshot, not a rule.
If the per-megapixel model fits your workload better, for instance very large outputs that you downscale anyway, nothing stops you from using it directly. If you want one wallet and the same request shape across FLUX, Seedream, Nano Banana and the rest, the flat line is the simpler thing to budget.
Sources
Related posts
More in Pricing
- Omni Flash 1.1 reference-image fee: Runway adds 1 credit, Sume none
Runway adds 1 credit per reference image on Gemini Omni Flash 1.1; Sume bills output seconds only. 8-second examples show which is cheaper at each resolution.
- A 60,000-character narration on Sume TTS: 3 jobs, $2.85
Gradium tags 221 voices for narration. For long scripts on Sume TTS, here is the 20,000 character cap, the 1,200 second limit and the cost of 3, 4 or 6 jobs.
- Grok Imagine 1.5 Lite at 2 cents vs 1.5 at 8 cents: what Sume bills
xAI lists Lite at $0.020 and 1.5 at $0.080 a second. Sume carries only the 1.5 row, image-to-video, 4-15 s, with its own list rate. Which line applies.
- Haiku 5.5 with US-only inference: 1.1x on both price tiers
Anthropic's pricing page applies the 1.1x US-only multiplier to Haiku 5.5's over-100k prices too. A worked agent-loop table and where Sume's own cap sits.
Written by Sume