Count the whole job: smarter budgets for AI images and tokens
One small rate can hide a large bill
A price per image or per million tokens is a starting point. Your job cost depends on how many outputs you need, their settings and how much work the model produces. Comparing the whole job helps you choose a seller that is affordable for your actual use.
Penny Owl shows image prices to three decimal places because fractions of a cent matter across many generations. Its Brain tables separate input from output because reading a prompt and writing an answer often have different rates.
Images: compare the same size and quality
Use the 1K, 2K and 4K columns to begin your comparison, then check the exact dimensions and quality in the quote context. A low-quality price cannot stand in for a medium-quality model setting. Reference images and editing may add charges. Some APIs bill for output tokens or megapixels rather than charging one flat amount.
For example, our October 5 check found an Atlas Cloud FLUX.2 Pro text-to-image quote of $0.030 at 1024 × 1024 and $0.075 at 2048 × 2048. One hundred images at those quoted settings would be $3.00 and $7.50. The 2K job costs 2.5 times as much in this example, not necessarily twice as much. Verify the current quote before using that estimate.
More pixels do not always mean a better model for your purpose. Drafts, small social graphics and print assets have different needs. Choose the smallest supported size that meets your goal, and account for unsuccessful creative attempts as well as the final images.
Brain: choose one seller for the entire request
The cheapest input rate and cheapest output rate can come from different sellers. You normally send a complete request to one provider, so adding those two unrelated minima can produce a bill you cannot actually obtain. Compare each seller's combined input and output cost.
The formula is: input tokens ÷ 1,000,000 × input rate, plus output tokens ÷ 1,000,000 × output rate. Here is a hypothetical example. Seller A charges $1 input and $10 output. Seller B charges $2 input and $3 output. With 100,000 input tokens and 10,000 output tokens, A costs $0.20 and B costs $0.23. With the same input and 100,000 output tokens, A costs $1.10 and B costs $0.50. The best seller changes with the job.
The job-budget calculator on Penny Owl's Brain model pages compares one seller's paired rates. It estimates standard uncached usage. Long-context thresholds, reasoning tokens, caching, batch discounts, tools, funding fees and taxes can alter the total; check the source and the model's limits.
Treat unknown prices honestly
A dash can mean unsupported size, missing public pricing, account-only access, a failed read or stale evidence. It does not prove the provider lacks the model. Consumer subscription credits and API dollars are not interchangeable without a documented conversion and matching settings.
Penny Owl is a free nonprofit community guide. We compare verified offers to help you spend less time searching and make a better choice. We do not hold your API keys or run your jobs from these public tables.