About & methodology

CostMyTokens exists because LLM API pricing has become genuinely hard to track: three major providers, frequent launches, promotional rates with expiry dates, speed tiers, batch tiers, caching discounts and long-context surcharges — while many calculators on the web quietly mix official prices with gateway reseller prices, or keep charging you with last season's numbers. We built the calculator we wanted to use ourselves.

How prices are verified

Every price in our dataset is taken from the provider's official pricing page and cross-checked against a second, independent source before it is published. Each model carries a verification date and a link to its source, shown on its page. The full dataset is re-checked on a weekly cycle, and new models are added within 48 hours of release once their official prices can be confirmed. When a provider announces a future price change — such as Gemini 4 Argon's rates doubling on December 1, 2026 — the change is noted on the model page in advance and applied on the day it takes effect.

Where a provider has not published a figure (for example, a cached-input rate), we show "Not published" rather than estimating one. An empty cell is honest; an invented number is not.

How costs are calculated

Cost per call = (input tokens × input price + output tokens × output price + cached tokens × cached price) ÷ 1,000,000. Periodic cost multiplies by your call volume. Token counting is provider-specific: OpenAI models are counted with the real o200k tokenizer in your browser; where a provider publishes no official browser tokenizer, the count is labeled an estimate. Your text is never uploaded — all calculation happens locally on your device.

Independence

Rankings and comparisons on this site are ordered by price and data only. If affiliate links are introduced in the future, they will be disclosed, and they will never change a price, a ranking, or a verdict.

Dataset last updated: 2026-10-24 · 10 models tracked across 3 providers.