Google's Gemini Decisions API Scores 1,000 Questions in One Pass

Google launched the Gemini Decisions API on November 11, 2026, an endpoint that evaluates as many as 1,000 independent questions across text, images, audio, video and PDFs in a single request. It runs on Gemini Decision 1 and is available through the Gemini API in Google AI Studio.

The endpoint targets bounded judgments — tool selection, policy checks, model routing, context grading and ranking — not prose or explanations. A request pairs one shared input with a set of independent questions. Instead of decoding answer text token by token, Gemini Decision 1 scores candidate options in one forward pass, so programs can branch, sort and route directly on typed responses.

Why skip the explanations?

The design addresses software control flow. Under the old pattern, a generated explanation would have to be converted back into structured values before a program could act. The Decisions API returns those structured values directly.

Each answer carries 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 and escalate uncertain ones. Google's documentation recommends setting thresholds with labeled examples from a developer's own traffic, and keeping a human or second check for high-stakes actions such as payouts.

What can developers ask?

The API's documentation sets a 262,144-token context window covering the shared input and all questions. Requests can mix three question types:

• Yes/no checks
• Fixed-set choices, accepting two to 255 options
• Ordered rubric scores, returned as probability-weighted averages of level indices with per-level probabilities

Questions can't see one another, and adding or removing one doesn't change the other answers. Dependent decisions belong in application logic: developers can send a second request after reading the first answer, or ask both questions together and discard the one they don't need.

How fast and how cheap?

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 stayed nearly flat between one and 100 questions per request. Gemini Decision 1 also led Google'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; each question adds its own instruction and criteria tokens. There are no variable thinking-token charges and no separate audio-transcription or 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.

Who's already using it?

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. CC from Google Labs uses it to select among more than 200 emoji reactions.

What are the current limits?

Developers can work in Python, JavaScript or REST; 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. Gemini in Google Cloud support is listed as coming soon.

The restrictions are real. Initial access covers only the Gemini Developer API v1beta with Gemini Decision 1. Question instructions are text-only. Conversation history and tool-call blocks are unsupported, and context caching, the Batch API, and Flex or Priority inference are unavailable.

For teams building agents, the API consolidates multimodal judgments into a single typed call. But responsibility for safe automation stays with the developer's criteria, labeled evaluation data and escalation rules — especially where an action carries financial or safety consequences.