Google releases Gemini Decisions API, which can evaluate 1,000 multimodal questions in a single request
On November 11, 2026, Google launched the Gemini Decisions API powered by Gemini Decision 1, available through the Gemini API in Google AI Studio, which can evaluate up to 1,000 independent questions spanning text, images, audio, video, and PDFs in a single request. The interface targets bounded judgments such as tool selection, policy checks, and model routing; the model scores candidate options in a single forward computation rather than generating text token by token, and each answer comes with a confidence score from 0 to 1. The documentation gives a 262,144-token context window, with pricing at $0.10 per 1 million input tokens and no charge for output tokens; Google says its median model compute latency is 50 to 100 milliseconds, and says that on its own benchmark suite it is nearly 5 times cheaper than Gemini 3.5 Flash-Lite paired with the Interactions API.
Why read this
SelectedAI score 78
The original text provides the pricing, latency, and confidence mechanism of the Gemini Decisions API, which can be used to judge how it replaces explanatory output to connect to control flow.
Topics
Body
Google on November 11, 2026 introduced the Gemini Decisions API, a new endpoint powered by Gemini Decision 1 and available through the Gemini API in Google AI Studio. It evaluates as many as 1,000 independent questions across text, images, audio, video and PDFs in one request.
The endpoint is built for bounded judgments such as tool selection, policy checks, model routing, context grading and ranking, rather than prose or explanations. A request combines one shared input with independent questions. Instead of decoding answer text token by token, Gemini Decision 1 scores candidate options in one forward pass, allowing programs to branch, sort and route directly on typed responses.
The approach is aimed at software control flow, where a generated explanation would otherwise have to be converted back into structured values before the program could act.
The Decisions API documentation sets a 262,144-token context window covering the shared input and all questions. Requests can mix yes/no checks, fixed-set choices and ordered rubric scores. Choice questions accept two to 255 options, while rubric answers are probability-weighted averages of level indices accompanied by per-level probabilities.
Every answer includes a confidence score from 0 to 1, defined as the estimated probability that the top answer is correct. Applications can act automatically on high-confidence results or escalate uncertain ones. The documentation recommends setting thresholds with labeled examples from a developer’s own traffic and retaining a human or second check for high-stakes actions such as payouts.
Questions cannot see one another, and adding or removing one does not change the other answers. Dependent decisions therefore belong in application logic: developers can send a second request after reading the first answer, or ask both questions together and disregard the answer they do not need.
Reported Latency, Pricing and Benchmarks
Google reports 50–100 milliseconds of median model-compute latency and 150–250 milliseconds of end-to-end latency across multimodal inputs and contexts up to 16K tokens. Latency remained nearly flat between one and 100 questions per request. Gemini Decision 1 also led the company’s benchmark suite spanning agent routing, reasoning, ranking and multimodal understanding, with the lowest calibration error.
Pricing is $0.10 per 1 million input tokens, with zero output tokens billed. The shared input is billed once per request regardless of question count, while each question adds its own instruction and criteria tokens. There are no variable thinking-token charges or separate audio-transcription and OCR costs. Google describes Gemini Decision 1 with the Decisions API as nearly five times cheaper across the shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API.
Google Deployments and Developer Access
Google says Gemini Notebook uses the API as a real-time router for model tiers and thinking levels, delivering up to 40% lower latency for everyday queries while reserving deeper reasoning models for complex tasks. Stitch’s rapid canvas experiment uses it to generate playable wireframes, while CC from Google Labs selects among more than 200 emoji reactions.
Developers can use Python, JavaScript or REST; the SDK quickstarts require google-genai 3.2 or later. A no-cost AI Studio tier supports prototyping. Integrations are rolling out across OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode, while Gemini in Google Cloud support is listed as coming soon.
Initial access is limited to the Gemini Developer API v1beta with Gemini Decision 1. Question instructions are text-only, and conversation history and tool-call blocks are unsupported. Context caching, Batch API, and Flex or Priority inference are also unavailable.
For teams building agents, the API consolidates multimodal judgments into a single typed call, but responsibility for safe automation remains with the developer’s criteria, labeled evaluation data and escalation rules, especially where an action carries financial or safety consequences.