Gemini Decisions API cost per 1,000 decisions. Next to a generative model doing the same job.

Cost per decision = input tokens x the input price per 1M / 1,000,000. On a generative model, add output tokens x its output price per 1M / 1,000,000. gemini-decision-1 generates 0 output tokens, so Google bills input tokens only. The output field applies to the generative models. Per 1,000 decisions = cost per decision x 1,000. Per month = cost per decision x decisions per month. One decision here is one request with one question. Several questions about the same input in one request count the shared input once; add each question's own instruction and criteria tokens to the input. Both sides use the same input token count. Tokenizers and prompt wording differ between models, so read usage from your own responses. Generative output includes thinking tokens: Google's page bills them as output. Prices: Google's paid tier, Standard tab, read 2026-11-11. The free tier, Batch, Flex and Priority are not used. The Decisions API supports none of Batch, Flex, Priority or context caching. Cost only. The calculator says nothing about whether two models give the same answers.

Google bills gemini-decision-1 per million input tokens only. It generates 0 output tokens, so output is never billed, and text, image, video, audio and PDF input share one price. The calculator above and the Gemini Decision 1 table on /models show the paid tier price as Google's pricing page lists it, read 2026-11-11.

Yes. Google's pricing page lists gemini-decision-1 input as free of charge on the free tier, in Google AI Studio and the Gemini API. The same page says free tier data is used to improve Google's products. Rate limits are on Google's rate limits page, linked from the model page.

Two reasons, both on Google's pricing page: gemini-decision-1 has a lower input price per million tokens than the Flash models compared here, and it writes no output. A generative model bills the label it writes back, and its thinking tokens, at its output price. The calculator shows the difference for your own token counts.

Only by the tokens of each extra question. Google's Decisions API guide says the shared input counts once per request, and each question adds its own instruction and criteria tokens. A request takes 1 to 1,000 questions within 262,144 input tokens.

Not on the Decisions API. Google's guide says it supports none of context caching, Batch API, Flex or Priority, and repeated input across requests counts in full. The Flash models have Batch and Flex prices on Google's page; this calculator uses the Standard price on every model.

The calculator compares cost only. Google's docs say gemini-decision-1 returns calibrated probabilities with a confidence for yes/no, choice and rubric questions. Google's guide says to set thresholds from labeled examples of your own traffic: test both models on them before you switch.

Yes. It runs in your browser, needs no account and makes no API call. Nothing you type is sent anywhere.