FLUX.2 Klein 9B from $0.015 vs Sume image prices for 4,000 images

Black Forest Labs lists FLUX.2 Klein 9B from $0.015 per image. Four thousand images at that starting price is $60.00. On Sume, which does not list any Klein model, the cheapest image rows for 4,000 images run from $99.00 (gpt-image-2.5, low quality, 1K) to $150.00 (flux-2-pro). That is a 1.65x to 2.5x difference.

Cost for 4,000 images

BFL's page says FLUX.2 prices are per megapixel, with the first megapixel at a flat rate and each extra megapixel adding to the total. The 'from' figure is therefore a floor for small outputs, not a quote for every size. Treat the $60 as a lower bound.

What the comparison does not say

A lower price does not make two models interchangeable. Klein is a smaller FLUX.2 variant; gpt-image-2.5 at low quality is a different model with different strengths, and Sume's low-quality row is a draft quality tier. The table only tells you the money.

On Sume the closest family member is flux-2-pro, at $0.0375 per image, which is the BFL 'from $0.03' Pro rate times 1.25. For 4,000 images that is $150.00.

When Sume's rows still make sense

If your pipeline already submits jobs to Sume and needs one wallet, one job format and one webhook shape, the extra $39 to $90 on a 4,000-image batch buys integration simplicity rather than a cheaper pixel. If you only generate small stills in bulk and have no other Sume work, the BFL rate wins on cost. See the Image API docs for what the Sume catalog lists.

A smaller batch for context

At 500 images the same rates give $7.50 for Klein 9B at its starting price, $12.38 for gpt-image-2.5 low on Sume (500 x $0.02475 = $12.375), and $18.75 for flux-2-pro. At that size the absolute gap is under $11, which is why the decision is usually about the model, not the price.

The gap grows with volume, and at 4,000 images the cheapest Sume row is $39 above the Klein starting price.