How we compare LLM API prices
Learn how AIPriceRank compares LLM API prices, estimates tokens for your budget, and calculates savings against official rates.
What your budget buys
Purchasing power is the number of input and output tokens your budget can buy at a provider's listed rates. It measures price alone, without assessing uptime or model quality.
Formula
weightedPricePerMillion = (inputRatio × inputPrice + outputRatio × outputPrice) ÷ (inputRatio + outputRatio) totalTokens = budget × 1,000,000 ÷ weightedPricePerMillion inputTokens = totalTokens × inputRatio ÷ (inputRatio + outputRatio) outputTokens = totalTokens × outputRatio ÷ (inputRatio + outputRatio) Providers are sorted by totalTokens, highest first. Missing or zero prices are excluded.
Worked example
Prices are USD per 1 million tokens. With a $100 budget and a balanced 1:1 mix, an input price of 2 and an output price of 6 give a weighted price of (2 + 6) ÷ 2 = 4. Total tokens = 100 × 1,000,000 ÷ 4 = 25 million, split evenly into 12.5 million input and 12.5 million output. The same model at an input price of 1 and an output price of 3 has a weighted price of 2 and buys 50 million tokens. The same $100 buys twice as many tokens.
Usage patterns
Scenarios include balanced conversation, quick questions, development, agent automation, document analysis and long context / RAG. Each option shows its input:output mix. Development includes code context; agent automation includes context sent repeatedly during multi-step tasks. These mixes are illustrative, not measured usage for any particular tool. Matching scenarios use the same mix in both calculator modes. Recorded cache prices are not part of the weighted price.
How we calculate savings
We apply the same budget and input/output mix to official rates. Savings = 1 − official token count ÷ provider token count. If your budget buys twice as many tokens through a provider, its rates are 50% lower for that mix. Equal token counts are shown as 1×. We show no savings percentage when official pricing is unavailable.
Homepage rankings and model comparisons
Homepage rankings use a $100 budget and an equal split of input and output tokens. Models are grouped by AI lab. Each savings percentage compares the provider's rates with official pricing; bars compare token counts with the top-ranked provider for that model. On a model page, adjust your budget and input/output mix to compare token estimates, or enter your usage to estimate per-request and monthly costs.
Data sources
Provider rates come from public pricing pages, and official rates come from the model developer. We record the source URL and collection date for each price. Comparisons use the latest available pricing data.