FLUX API per-image price: what 500 images cost at $0.024 to $0.048

Black Forest Labs lists FLUX API pricing at roughly $0.024 to $0.048 per image depending on model and resolution, billed pay-as-you-go with no subscription. At those rates 500 images cost between $12.00 and $24.00 before retries.

What the pricing page says

The BFL pricing page describes a pay-as-you-go model with no subscriptions and a per-image price that starts at $0.024 and reaches $0.048 across models and resolutions. It does not give a single price for every case, so a plan has to carry both ends of the range.

Pay-as-you-go changes how you plan

With no subscription there is no monthly floor to amortize, so the cost of an idle month is zero and the cost of a busy month is linear in image count. That makes campaigns easy to budget: multiply expected keepers by the inverse of your keep rate, then by the per-image price.

For example, if you keep one image in four, 500 keepers need 2,000 generations. At the low end of the range that is $48.00, and at the high end $96.00. A pilot of 40 images will tell you your real keep rate before you commit to the full run.

Checking the price you will actually pay on Sume

Vendor list prices are a guide, not your invoice. On Sume, the endpoint pricing line (GET /v1/images/models/{model_id}/endpoints) is the amount charged to your wallet, and cost_usd × n is what a request costs. Every finished response includes usage.cost. Generation is all-or-nothing, so a failed image is not billed and does not add to the retry budget.

Because Sume's n range is per model, read the n descriptor before you ask for a large batch in one call. A request above the model's ceiling is a validation error, not a partial result.

curl "https://api.sume.com/v1/images/models" \
  -H "Authorization: Bearer $SUME_API_KEY"
# read each model's supported_parameters.n.max before setting n

A short checklist

• Fix the output resolution first, since it moves the price within the range.
• Run a 40-image pilot to measure your keep rate.
• Multiply by 1 over keep rate, then by both ends of the range.
• Compare against the Sume endpoint line for the same model before you commit.