Decision guide · API pricing
How to compare AI model API prices.
The useful number is the cost of your workload at a particular provider, using a dated price quote. A single input-token price rarely tells the whole story.
Start with your workload
Write down monthly requests, average input tokens per request, and average output tokens per request. For paired text-token rates, the basic projection is:
For example, 100,000 requests with 1,200 input and 350 output tokens at hypothetical rates of $1 and $4 per million tokens produce a $260 monthly projection. This is a calculation example, not a quoted model price or verified saving.
Compare provider offerings, not just model names
One model can have multiple provider offerings. Each can differ in price and reported limits. ALM keeps the provider attached to each rate and lets you change the provider in the comparison tool. Open the model page to inspect its source record and observation date before using a quote in a purchasing decision.
Know what the estimate leaves out
Text-token estimates do not automatically include image or audio charges, cache behavior, extra reasoning tokens, request fees, retries, discounts, or negotiated terms. The advanced calculator exposes more of these inputs. If a needed price is missing or stale, treat the cost as unavailable rather than zero.
Price is one constraint
Check required input and output modalities, tool support, context, and provider availability before comparing cost. A lower projected bill does not establish equal quality for your documents, code, or other workload. ALM’s feature index is a documented heuristic, not a task-quality score. For consequential choices, test the candidates on representative examples of your own work.
How ALM presents prices
The directory shows a reference offering to make scanning easier. The comparison tool calculates with the selected provider’s eligible paired text-token rates. ALM records source provenance and retains price history. Read the methodology for quote selection, freshness rules, and formula versions; inspect data sources for licensing and attribution.