[KongchangAI]
ModelGemini Decision 1

Gemini Decision 1

A multimodal decision model introduced by Google, positioned to make judgments rather than generate content. It supports text, image, audio, video, and PDF input and outputs boolean probabilities, options, and scores, without consuming output tokens.

Core Facts

Timeline (last 90 days)

Nov 14

Gemini Decision 1, the underlying model of the Gemini Decisions API, supports uploading the same material only once to answer 1,000 independent questions; the interface bills by input, charges no output token fees, and input costs about $0.1 per million tokens.

Unverified50%
Nov 13

Google launched the decision model Gemini Decision 1, which supports audio modality input, whereas OpenAI's corresponding product and the local Ollama solution do not support audio modality.

Unverified50%
Nov 13

In the author's email classification comparison, the latency of each solution was approximately: about 1.2 seconds for the local model after warm start, about 2 seconds for Jeff, about 3.4 seconds for Gemini, and about 6.5 seconds for GPT.

Unverified50%
Nov 13

According to Artificial Analysis's decision model index, Gemini Decision 1 performs roughly on par with Jeff and better than GPT-6 Luna Decision and Cloudflare's Clef series.

Unverified50%
Nov 11

Gemini Decision 1, Jev and OpenAI have formed a three-way competitive landscape in the decision model field.

Unverified50%
Nov 11

Gemini Decision 1 supports three types of questions: yes/no judgments, multiple-choice selection and rating-scale scoring.

Unverified50%
Nov 11

Gemini Decision 1 and the GPT-6 Luna decision model are priced the same: 10 cents per million input tokens, with output free.

Unverified50%
Nov 11

Jev supports only text and JSON input, OpenAI supports text and images, while Gemini Decision 1 covers text, images, audio, video and PDF.

Unverified50%
Nov 11

Gemini Decision 1 returns probability values rather than hard binary results for yes/no questions; in the test, dragon was 0.0498, France was 0.0833, robot was 0.9234 and United States was 0.9230.

Unverified50%
Nov 11

The probabilistic output design allows downstream systems to set thresholds for graded processing, but callers need to handle the threshold logic themselves and cannot directly use 0/1 boolean values.

Unverified50%

9 more timeline events

All Facts (19)

Partially Verified

Google's Gemini Decision 1 model launched on AI Gateway, positioned as a decision model rather than a content generation model.

70%
Unverified

Gemini Decision 1, the underlying model of the Gemini Decisions API, supports uploading the same material only once to answer 1,000 independent questions; the interface bills by input, charges no output token fees, and input costs about $0.1 per million tokens.

50%
Unverified

In the author's email classification comparison, the latency of each solution was approximately: about 1.2 seconds for the local model after warm start, about 2 seconds for Jeff, about 3.4 seconds for Gemini, and about 6.5 seconds for GPT.

50%
Unverified

According to Artificial Analysis's decision model index, Gemini Decision 1 performs roughly on par with Jeff and better than GPT-6 Luna Decision and Cloudflare's Clef series.

50%
Unverified

The probabilistic output design allows downstream systems to set thresholds for graded processing, but callers need to handle the threshold logic themselves and cannot directly use 0/1 boolean values.

50%
Unverified

Jev supports only text and JSON input, OpenAI supports text and images, while Gemini Decision 1 covers text, images, audio, video and PDF.

50%
Unverified

Gemini Decision 1 returns probability values rather than hard binary results for yes/no questions; in the test, dragon was 0.0498, France was 0.0833, robot was 0.9234 and United States was 0.9230.

50%
Unverified

Gemini Decision 1 supports three types of questions: yes/no judgments, multiple-choice selection and rating-scale scoring.

50%
Unverified

Gemini Decision 1 and the GPT-6 Luna decision model are priced the same: 10 cents per million input tokens, with output free.

50%
Unverified

Gemini Decision 1, Jev and OpenAI have formed a three-way competitive landscape in the decision model field.

50%
Unverified

When official benchmark data is lacking, developers should build their own small-scale controlled tests, running candidate models on the same set of branching decision questions and recording judgment consistency and structured-output parsing success rates.

50%
Unverified

At this stage, Gemini Decision 1 should not directly replace models in production environments, but should be treated as a candidate pending verification.

50%
Unverified

Gemini Decision 1 supports text, image, audio, video, and PDF document inputs, and outputs boolean probabilities, choices, and scores, without consuming output tokens.

50%
Unverified

In the problem definition object, each question has a type (boolean, choice, score) and instructions, and the model returns the corresponding answer fields as declared; this structure is enforced at the API level rather than relying on prompts.

50%
Unverified

PDF, audio, and video inputs are currently supported only through AI SDK 7; the TypeSafe-compatible API and the Decisions API support only text and image inputs.

50%
Unverified

Decision calls appear in AI Gateway logs like other model requests and are counted against the budget.

50%
Unverified

The decision model outputs only probabilities, options, and scores, with no explanations; when a business requires an auditable chain of reasoning, a generative model or an additional explanation layer is still needed.

50%
Unverified

Output token consumption is tied to the model's token-by-token autoregressive decoding; a decision model constrains the output space to a finite set of answers (booleans, options, or scores), thereby bypassing token-by-token decoding.

50%
Unverified

Google launched the decision model Gemini Decision 1, which supports audio modality input, whereas OpenAI's corresponding product and the local Ollama solution do not support audio modality.

50%

Source Articles