THE APEX TIMES
The new Gemini Decisions API is designed to score bounded yes/no, multiple-choice, and rubric questions across text, images, audio, video, and PDFs in a single pass, with confidence scores and pricing aimed at low-latency decisioning.
Google is rolling out a new Gemini Decisions API aimed at making “bounded” AI decisions faster and cheaper for developers. Announced in a Google Blog post on November 11, the product is built on Google’s Gemini Decision 1 model, which scores candidate options rather than generating free-form answers.
In practical terms, Google says the Decisions API can evaluate up to 1,000 independent questions against a shared multimodal input in one request. The shared input can combine multiple forms of data, including text plus images, audio, video, and PDFs, and is limited to 256K tokens. The API returns probabilities and confidence scores for each question, giving applications a announcement for when the system is confident versus when it should route to a fallback workflow such as a human review.
Google positions the endpoint as a low-latency “control plane” for common decisioning tasks. The blog post lists uses such as tool selection, policy checks, model routing, and context grading. Instead of token-by-token decoding, Gemini Decision 1 is described as performing a single forward-pass scoring step over the supplied options, which Google says helps keep runtime consistent and fast.
The company also gave performance and calibration details. Google says the end-to-end latency is in the 150 to 250 ms range, and that median model compute latency remains within a 50 to 100 ms band for the scoring step. It also says calibration error is the lowest across benchmark suites covering agent routing, reasoning, ranking, and multimodal understanding.
Pricing is a central part of the pitch. Google states Gemini Decision 1 for the Decisions API is $0.10 per 1M input tokens and $0 for output tokens. It also says the input is billed once per request regardless of the number of questions evaluated, with each additional question adding only its instruction and criteria tokens. Google further claims there are no separate output-token charges, no variable “thinking” token billing, and no additional costs for transcription and OCR inside the workflow.
The blog post explains that developers can mix question types in a single call, including yes/no checks, multiple-choice questions, and rubric scoring. Each answer includes a confidence score from 0 to 1, intended to let code automatically route ambiguous cases to other systems. Google also says the approach is built on Gemini’s native multimodal architecture, allowing text, images, audio, video, and PDFs to be processed directly within one shared input rather than handled through separate preprocessing steps.
Google describes the API as designed for routing and grading use cases where developers want probabilistic outputs without the overhead of generating an answer essay. It gives an example of checking a photo of a breakroom counter together with a text note against different question types within one request. It also notes that teams across Google are using the Decisions API as a low-latency control plane by defining options rather than training custom classifiers or waiting on slower generative calls.
The rollout includes developer tooling and ecosystem integrations. Google says users can start building by getting an API key in Google AI Studio and reviewing documentation for Python, JavaScript, and REST. It also references demo applications such as Doorstep Parcels, Speed Sketch, and Find the Zoo Keeper. Google adds that integrations are rolling out across OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit, and OpenCode, along with planned support for Gemini in Google Cloud.
Still, some practical details remain unstated in the announcement. Google does not specify what kinds of rubric formats are supported beyond describing scoring, nor does it disclose how confidence scores are calibrated in edge cases where inputs are noisy or heavily truncated. It also does not provide a public, independently audited benchmark methodology in the post, even though it cites leading calibration performance across test suites.
A Wall Street note captured a split view of Advanced Micro Devices, with higher profit expectations offset by a lower price target amid recent market pressure.
A Yahoo Finance report says Cathie Wood’s ARK Investment Management reduced its Amazon position while adding to a basket of aerospace-related names after strong trading activity around a rocket launch company.
The browser-based retro video game debuts on Googlebook, with Google pledging $11 per play to Movember men’s health causes through November 30, up to $500,000.
An equity note circulated by Yahoo Finance points to AI chip testing and a widening Nvidia relationship as key catalysts for Teradyne, the maker of semiconductor test equipment.
The comments, reported by Yahoo Finance, put a spotlight on how cloud AI infrastructure providers are weighing Nvidia’s role as scrutiny continues around the broader semiconductor push for artificial intelligence.
In a portfolio-related update cited by Yahoo Finance, the veteran investor said he was positioning for a near-term lift in Nvidia after the stock’s recent selloff, while also questioning the durability of the company’s competitive advantage.
A market report highlights how Anthropic’s momentum in generative AI could reshape competitive dynamics for Microsoft’s cloud and software stack, adding pressure to an already crowded field.
Nvidia and AMD are coming off record AI-driven quarters, yet a new political push to restrict where advanced semiconductors can be sold is raising questions about how much growth either company can sustain as the year turns.
Hilary Maxson says Oracle expects to generate free cash flow at a higher clip, including a promise of more than $30 billion in fiscal 2029 and a return to cash-flow positivity in the second half of fiscal 2028, following a roughly 70% share price decline.
A market commentary argues that Microsoft’s continued buildout of data-center capacity is being misread as unproductive cash burn, and that a less-discussed line item on the company’s balance sheet can alter how investors judge the spending.
Comments
Community guidelines. Comments may be removed for violations.
Loading comments...