Author|Dongmei (AI Frontline)

While the entire AI industry is still competing over who can generate more tokens and who has stronger reasoning abilities, Google has launched a new model that does not rely on word-for-word generation of answers.

Just now, Google officially released Gemini Decision 1, along with the Gemini Decisions API powered by this model. Unlike traditional generative large models, this model is specifically designed for rapid decision-making without needing to generate text token by token but instead scores pre-set candidate answers through a single forward pass.

Google CEO Sundar Pichai also posted on the X platform to introduce the main capabilities of Gemini Decision 1. He wrote:

Today, we officially released the Gemini Decisions API in Google AI Studio, supported by the brand new Gemini Decision 1 model.

Unlike traditional models that generate text token by token, this model is specifically designed for decision tasks within a closed set of options. It can score up to 1000 questions from text, images, audio, video, and PDFs in a single processing pass.

Here are a few noteworthy details:

· Supports three types of decision-making: Yes/No judgments, multiple-choice (up to 255 options), and scoring based on criteria;

· The price is only $0.10 per million input tokens, and the output token cost is zero, about one-fifth of Flash-Lite's cost;

· Leads in multiple benchmark assessments, with the lowest calibration error.

Google's internal teams have already begun using this technology.

In the past, when developers wanted AI to determine whether an email was spam or to decide which agent should handle a task, they often had to call a generative large model and then extract results from the generated text. Gemini Decision 1 eliminates the need for word-by-word decoding.

Developers only need to provide input content, the question to be judged, and candidate options, and the model can directly return the corresponding judgment results, probabilities, and confidence levels.

This means that for tasks that only require a classification, scoring, or selection result, AI no longer needs to complete a full text generation process.

According to data released by Google, Gemini Decision 1 can handle up to 1000 independent questions in a single request, supporting multi-modal inputs such as text, images, audio, video, and PDFs, with an end-to-end latency of only 150-250 milliseconds.

The pricing is also relatively aggressive: only $0.10 per million input tokens, with no charge for output tokens. In the same set of benchmarks released by Google, its invocation cost is nearly five times lower than Gemini 3.5 Flash-Lite.

More importantly, this is not just a model for answering multiple-choice questions. Google has already used it in scenarios such as AI model routing, agent tool selection, product design, and security detection, attempting to make it a foundational component specifically responsible for judgment and decision-making in AI systems.

Process 1000 questions at once, up to 255 options

According to the technical introduction released by Google, Gemini Decision 1 provides three core decision-making capabilities.

The first type is Yes/No (binary judgment). For example, determining whether there is a security hazard in an image, or whether a piece of text violates established rules. The model will give the probability of the respective conditions being met.

The second type is Multiple Choice. Developers can preset 2-255 candidate options. The model will not only select the most suitable answer but will also return the normalized probabilities for each candidate option.

The third type is Ordered Rubric (ordinal scoring). Developers can define different levels of scoring criteria for the model to evaluate the input content and return continuous weighted scores and probabilities at each level.

The three tasks can also be mixed together in a single request.

Google provided a simple example: upload a photo of an office kitchenette counter to the model and attach a text description, asking it to simultaneously determine whether there is liquid spillage on the countertop, what cleaning measures should be taken, and the severity of the situation.

The traditional approach may require designing multiple classifiers or executing different prompts using generative large models. Gemini Decision 1 can complete these judgments in a single request.

Additionally, the model supports multi-modal input contexts of up to 256K tokens. For enterprise scenarios, this means developers can analyze call recordings alongside customer data or combine insurance claim documents with onsite photos to make judgments.

Each decision will provide a confidence level from 0 to 1, making it easy for the system to refer uncertain results to other models or for human review.

Another competitive advantage of Gemini Decision 1 is its speed and cost.

According to Google's official announcement, in benchmark evaluations covering agent routing, reasoning, ranking, and multi-modal understanding, Gemini Decision 1 performed ahead of others and achieved the lowest calibration error.

In test contexts not exceeding 16K tokens, its model computation stage median latency remains at 50-100 milliseconds, with an end-to-end delay of about 150-250 milliseconds. Even when the number of questions in a single request increases from 1 to 100, the latency remains stable.

On the common benchmark test set used by officials, the cost of the Gemini Decisions API is nearly five times lower than calling Gemini 3.5 Flash-Lite via the Interactions API.

7 first places out of 9 tests with a maximum accuracy of 92.2%

In this release, Google also announced a noteworthy comparative scorecard for the models.

Instead of comparing with generally large models as usual, Google selected three competing products this time: OpenAI's GPT-6 Luna, the specialized classification decision model TypeSafe Jev 1.14, and its own Gemini 3.5 Flash-Lite.

The results show that Gemini Decision 1 took first place in 7 out of 9 benchmark tests, leading in probability calibration as well, although in two other tests and total invocation costs, competitors performed better.

In the JevBench test, Gemini Decision 1 achieved an accuracy of 92.2%, surpassing GPT-6 Luna's 90.9%, TypeSafe Jev 1.14's 90.0%, and Gemini 3.5 Flash-Lite's 86.6%.

In the Amazon ESCI product search relevance test, Gemini Decision 1 scored 58.7% in the Macro-F1 score, 4 percentage points higher than GPT-6 Luna and 11.5 percentage points higher than Flash-Lite.

In the New Yorker Caption test, the gap widened further. Gemini Decision 1 scored 79.7%, while GPT-6 Luna scored 68.4%, a difference of 11.3 percentage points. This test involved understanding and selecting captions for cartoons, testing the model's contextual understanding and candidate content judgment capabilities.

In the medical question-answering test PubMedQA, Gemini Decision 1 also ranked first with a score of 79.8%, higher than GPT-6 Luna's 74.4%.

Similar results were seen in the multilingual test Global-MMLU-Lite. Gemini Decision 1 achieved 86.0%, while GPT-6 Luna scored 79.6%, a difference of 6.4 percentage points.

However, Google did not achieve a comprehensive lead.

In the Mind2Web Multimodal test, TypeSafe Jev 1.14 ranked first with a score of 86.8%, followed by Gemini Decision 1 at 86.2%, trailing by 0.6 percentage points.

In the When2Call test, which assesses tool invocation judgment capabilities, TypeSafe Jev 1.14's advantage was even more evident, scoring 82.0%, while Gemini Decision 1 scored 79.9%.

This result is particularly noteworthy. Tool selection and timing judgment is one of the main application scenarios for Gemini Decision 1 aimed at agents, yet in this test, the specialized decision model TypeSafe Jev still maintains its lead.

Additionally, in the audio understanding test MMAU and the video understanding test Video-MME, Gemini Decision 1 scored 71.8% and 69.1%, respectively, both higher than Flash-Lite.

More noteworthy than accuracy is the probability calibration capability that Google emphasized this time.

In the Top-1 ECE (Expected Calibration Error) test, Gemini Decision 1's calibration error was only 0.0250, while GPT-6 Luna's was 0.0721, TypeSafe Jev 1.14's was 0.0494, and Flash-Lite reached 0.2104.

In other words, in this set of evaluations, Gemini Decision 1's calibration error was about 65% lower than GPT-6 Luna's and about 88% lower than Flash-Lite's.

This metric is particularly important for agent systems.

For example, when an agent needs to determine whether a piece of code has a security vulnerability, the model not only provides a result of "exists" or "does not exist," but may also include a probability value. The system can use the relationship between the probability and actual accuracy rate to decide whether to take action automatically or refer it for human review.

Google is already using it: Agent routing delay reduced by up to 40%

Unlike some new models that only showcase benchmark scores, Google also disclosed application cases of the Gemini Decisions API in its internal products.

Among the most noteworthy is Gemini Notebook.

Google is using the Decisions API as a real-time model router, selecting appropriate model tiers and reasoning intensity based on the complexity of user requests. For everyday simple questions, the system can prioritize calling lower-cost, faster-responding models; for complex reasoning tasks, requests can be handed over to more capable models.

Google stated that through this dynamic routing approach, the latency for everyday queries can be reduced by up to 40%.

This design is closely related to the current trends in Agent systems. When an agent completes a task, it often needs to continuously judge the next step, such as which tool to choose, which model to invoke, or whether manual intervention is required.

If every simple decision has to be handled by a large generative model, the system incurs extra delays and costs. A specialized decision-making model offers another choice: letting relatively simple, clearly defined judgments be handled by low-latency models and reserving complex tasks for general large models.

In addition to Gemini Notebook, Google disclosed three other application directions.

In Google Stitch, the newly launched Rapid Canvas experimental feature utilizes the Decisions API to create interactive product flow wireframes within seconds, rapidly responding to modifications issued by users through voice or clicks. Once users confirm the design direction, they can transfer the page into the complete Stitch workflow for further refinement.

In Google Labs' CC product, the Decisions API is used to quickly select appropriate emoji reactions from over 200 options. Within Google Search, Google has also used this API to build a security detection layer for the MCP tool, identifying potential adversarial behaviors and content.

Reference link: Gemini Decisions API Official Blog

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到[email protected]，本平台相关工作人员将会进行核查。