pricinggooglegemininew-models

Gemini Decision 1: $0.10 per Million Input Tokens, No Output Bill, and It Does Not Write Text

Short answer. Gemini Decision 1, released on November 11, 2026, costs $0.10 per million input tokens, and Google bills no output: it returns 0 output tokens. One request of 10,000 tokens with 50 questions costs $0.001. A million requests of 2,000 tokens a month cost $200.00. It answers yes/no, choice and score questions; it does not write text.

In plain words. This model does not chat or write. It reads what you send and answers fixed questions with probabilities. You pay only for what you send it. A token is a short piece of text, about three-quarters of a word.

What Google released

The Gemini API release notes carry an entry dated November 11, 2026: “Released Gemini Decision ( gemini-decision-1 ) and the Decisions API for fast, calibrated binary, categorical, and rubric decisions across text, images, audio, video, and PDFs with zero output tokens.” Google’s blog post of the same day says the model is “available through the Gemini API in Google AI Studio”.

The release note says “Released”. It does not say “generally available (GA)”, as the note for Gemini 4 Carbon did the day before, and it does not say preview either. The model page lists a single version, “gemini-decision-1”, with no “Stable” or “Preview” label, so we call it released and nothing more.

What it costs

Google’s pricing page, read on November 12, 2026, has a block for “Gemini Decision gemini-decision-1” with three rows that matter:

Row on the pricing pageFree tierPaid tier, per 1M tokens
Input (text, image, video, audio, PDF)Free of charge$0.10
OutputNot applicable (0 output tokens)Not applicable (0 output tokens)
Context cachingNot availableNot available

The literal rows are “Input price (text, image, video, audio, PDF) | Free of charge | $0.10 |” and “Output price | Not applicable (0 output tokens) | Not applicable (0 output tokens) |”. Above the table the page says: “Each request bills shared input tokens once, regardless of question count.” It adds: “Batch API isn’t supported.”

The blog post gives the same price in words: “Gemini Decision 1 is priced at $0.10 per 1M input tokens and $0 for output tokens, with an API no-cost tier you can use in Google AI Studio for prototyping.” On the shared input it says: “Your shared input is billed once per request no matter how many questions you evaluate (each question adds only its own instruction and criteria tokens).” The Decisions API guide puts it in five words: “Only input tokens are billed”.

What it returns, and what it is not

The guide is direct about the output: “It generates zero text tokens, so output token usage is always 0.” Each question has one of three types. A yes/no question returns “noul (0.01 to 0.99), confidence”. A choice question returns “choice , probabilities , confidence”. A score question returns “score , probabilities , legend , confidence”.

So Gemini Decision 1 is not a chat model, and it does not replace one. The model page lists “Text generation Not supported”, “Function calling Not supported” and “Thinking Not supported”. The guide’s own comparison sends “Generated text, summaries, or explanations” to the Interactions API, Google’s endpoint for its generating models. What Decision 1 does is answer closed questions about an input: is this photo damaged, which queue does this message go to, how does this answer rate on a rubric. A chat model is still needed to write the reply.

Limits

The guide’s limits table reads “Questions per request | 1 to 1,000 (bounded by the 262,144-token context window) |” and “Context window | 262,144 tokens across shared input and all questions |”. The model page gives “Input token limit 262,144” and “Output token limit 0 (up to 1,000 questions per request)”. There is no discount for sending the same input twice: “The Decisions API doesn’t support context caching, Batch API , or Flex or Priority inference. Repeated input across requests counts in full.”

On the free tier, the rate limits page prints “Free | 200 | 50,000 | 10,000 |”: 200 requests per minute, 50,000 tokens per minute and 10,000 requests per day.

Two requests, priced

Both examples are our own arithmetic from the $0.10 price. The token counts include the shared input and every question’s instructions and criteria, since Google says each question “adds only its own instruction and criteria tokens”. Output is 0 in both.

ExampleInput tokens billedGemini Decision 1
One request of 10,000 tokens with 50 questions10,000$0.001
1,000,000 requests a month of 2,000 tokens each2,000,000,000$200.00

The 50 questions do not multiply the bill: the shared input is billed once, and the questions only add their own tokens inside the 10,000. At a million requests a month, every token of shared input that can be cut saves money, because nothing is cached.

Google’s “nearly 5x cheaper” against Gemini 3.5 Flash-Lite

Google’s blog post says: “With zero output tokens billed, Gemini Decision 1 with the Decisions API is nearly 5x cheaper across the shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API.”

We can check the input part of that. Our row for Gemini 3.5 Flash-Lite has $0.30 per million input tokens, verified on November 9, 2026, and the pricing page read on November 12 still prints “$0.30 (text / image / video / audio)”. Against $0.10, that is 3×: Flash-Lite costs three times more per input token. The same 10,000-token request costs $0.003 in input on Flash-Lite.

The rest of Google’s “nearly 5x” comes from output. Flash-Lite also bills “Output price (including thinking tokens)” at $2.50 per million, and a generating model has to write its answer. The blog post measures the whole bill “across the shared benchmark suite” and does not print the token counts behind it, so we cannot recompute the 5x. On input alone the gap is 3×.

Where it is sold

The guide’s limitations start with “Available on the Gemini Developer API (v1beta) with Gemini Decision 1 only.” The blog post adds “support coming soon to Gemini in Google Cloud”. It is not yet offered through Google Cloud.

OpenRouter lists it too. Its page, read on November 12, 2026, says “Gemini Decision 1 costs $0.10/M input tokens and $0.00/M output tokens.” and names one provider, “Google AI Studio”, with a 262,144 token context. That is the same input price as Google’s page, resold.

Free tier and your data

The pricing page prints “Free of charge” for input on the free tier. The same block ends with the row “Used to improve our products | Yes | No |”: Yes on the free tier, No on the paid tier. Prototyping on the free tier means Google may use what you send to improve its products.

Why it is not in our pricing table

Our Google API pricing page, the calculator and the rankings compare input and output prices per token for models that write text. A model with no output price would look free on the output side there, which it is not: it simply produces no text. As of November 12, 2026, Gemini Decision 1 is covered in this post and not in those tables. Every other Google price is on that page.

How we checked

Every figure and sentence above comes from Google’s own pages, read on November 12, 2026: the Gemini API pricing page, the Decisions API guide, the Gemini Decision 1 model page, the rate limits page and the release notes, each through the Markdown version Google serves, and Google’s blog post of November 11, 2026. The new-models scan that AI Signal runs every morning flagged “gemini-decision-1” on the pricing page on November 12. OpenRouter’s model page and its endpoints API were read the same day. The Flash-Lite price is our dataset row, checked against the same reading of the pricing page. The request costs and the 3× ratio are our own arithmetic from the per-million prices, in US dollars, on the paid tier.

Frequently asked questions

How much does Gemini Decision 1 cost?

$0.10 per million input tokens on Google's pricing page read on November 12, 2026, which prints 'Not applicable (0 output tokens)' for output. A request of 10,000 tokens costs $0.001, and a million requests of 2,000 tokens cost $200.00.

Is Gemini Decision 1 a chat model?

No: Google's Decisions API guide says it 'generates zero text tokens, so output token usage is always 0' and returns yes/no, choice and score answers with probabilities. It does not write text, so it works next to a chat model, not in place of one.

Is Gemini Decision 1 free to try?

Yes, on the free tier: the pricing page prints 'Free of charge' for input, and its 'Used to improve our products' row reads Yes on the free tier and No on the paid tier. Google's blog post calls it 'an API no-cost tier you can use in Google AI Studio for prototyping'.

Drafted with AI assistance from the cited source and reviewed by Álvaro Lucero Rufino, who is responsible for what stays published here. Automated publishing was discontinued in September 2026; posts that no longer met that standard were removed rather than left up. This is original explanatory writing, not a reproduction — read the source above for the full story.