Google has launched Gemini Decisions API, which is a new endpoint driven by its Gemini Decision1 model. It can help AI intelligents make quick, preset decisions without the need to generate lengthy text responses.
This interface allows developers to evaluate up to 1000 independent questions for shared multimodal inputs that contain text, images, audio, video, and PDF. Gemini Decision1 does not generate responses for each individual token, but rather scores preset options in a single forward pass, returning probabilities and confidence scores. The end-to-end latency is between 150 and 250 milliseconds.
Google stated that this model is specifically built for bounded tasks, such as selecting tools, checking policy compliance, routing queries to the correct models, and evaluating contexts. Developers can submit "yes or no" checks, multiple-choice questions with up to 255 options, or scoring requests based on preset scoring criteria in a single API call.
Faster, lower-cost decision-making
According to Google, the median model computation time for this API is between 50 and 100 milliseconds, and the performance remains almost constant when each request processes 1 to 100 questions.
This model supports a context window of up to 256,000 token, and is capable of processing multiple types of data simultaneously. This enables developers to combine inputs such as audio recordings, images, and documents without relying on separate processing systems.
Google also claims that in benchmark tests covering agent routing, reasoning, sorting, and multimodal understanding, the model outperformed competing systems while achieving the lowest calibration error.
The company stated that in their shared benchmark testing suite, Gemini Decision1 is nearly 5 times cheaper than Gemini which uses Interactions API.
This release comes at a time when the competition among professional AI decision-making tools is intensifying. OpenAI launched its Decisions API in October, which is driven by GPT-6 and Luna models. The company claims that this service makes decisions 10 times faster than its Responses API and supports up to 400 questions per request. Similar to Google's products, it offers probabilistic checks, multiple-choice options, and scoring based on evaluation criteria, with pricing starting from $0.10 per million inputs.
Decisions API For every million inputs, token a fee of 0.10 US dollars is charged, with no fee for output token. Regardless of the number of questions assessed, developers only need to pay once for sharing input data, although each question will increase the token related to instructions and standards.
AI applications that empower the real world
Google has already deployed this technology in multiple products, demonstrating how dedicated decision-making models can improve the speed and efficiency of automated systems.
In Gemini Notebook, API routes the incoming queries to the appropriate model level and inference stage. According to Google, this method can reduce the latency of everyday queries by up to 40%, while reserving more powerful inference models for complex tasks.
This technology also provides support for Google's Stitch platform's rapid design experiments, allowing users to generate interactive product wireframes and make changes through voice commands or touch input.
At the same time, Google's CC assistant uses this API to select an emoji response from over 200 options in a fraction of a second.
Google Search is also using this technology as a security layer to detect adversarial behaviors involving model context protocol (MCP) tools.
In addition to these applications, this API can also help developers build systems by automatically selecting the appropriate tools, evaluating incoming information, and determining when human intervention is needed.
Decisions API is now provided through Google AI Studio, which includes a free tier for prototyping. Google is also launching integrations with developer platforms such as OpenRouter, LangChain, Vercel, LiteLLM, and PydanticAI.
