Back home

Gemini Decision 1 / Gemini Decisions API

sitemap+x+web 0.96
Visit source

Intro

Gemini Decision 1 is the new model Google announced on November 11, 2026, and Gemini Decisions API is the endpoint that serves it. The official post is Gemini Decisions API: Fast, calibrated multimodal decisions, by Min Choi, Product Manager, Google DeepMind, and Felix Yu, Principal Scientist, Google DeepMind. The blog says the company is introducing the Gemini Decisions API, a new endpoint for fast, closed-set decisions powered by the new Gemini Decision 1 model (gemini-decision-1), available through the Gemini API in Google AI Studio.

Officially, the Decisions API gives purpose-built primitives for bounded choices such as tool selection, policy checks, model routing, and context grading. You can pass one multimodal input across text, images, audio, video, and PDFs (up to 256K tokens) along with up to 1,000 independent questions. Rather than generating answer text token by token, Gemini Decision 1 scores candidate options in a single forward pass and returns probabilities and confidence scores. The blog gives end-to-end latency of 150–250 ms, and pricing of $0.10 per 1M input tokens with zero output tokens billed. It also says that, across a shared benchmark suite, Gemini Decision 1 with the Decisions API is nearly 5x cheaper than Gemini 3.5 Flash-Lite with the Interactions API, and that median model-compute latency is 50–100 ms (150–250 ms end-to-end) and remains nearly flat from 1 to 100 questions per request at contexts up to 16K tokens. Those latency, price-comparison and “leads the suite / lowest calibration error” figures are the company’s; this page does not independently reproduce them.

The official API has three question types: noul (yes/no, with optional true/false criteria), choice (2 to 255 options, returning the winner and normalized probabilities), and score (an ordered rubric, returning a continuous weighted score and per-level probabilities). Every answer includes a confidence score from 0 to 1 so code can send ambiguous cases to a fallback or a human reviewer. The input field uses the same Content block format as the Interactions API. Google says its own teams are already using it: Gemini Notebook as a real-time router for model tier and thinking level (up to 40% lower latency on everyday queries, with deeper models reserved for complex tasks); the rapid canvas experiment in Stitch, to generate playable product-flow wireframes in seconds; CC from Google Labs, to pick an emoji from a palette of more than 200 options; and Google Search, as a safety layer for MCP tools against adversarial evals within the latency envelope.

The blog says you can start today with an API key in Google AI Studio, including an API no-cost tier for prototyping, plus Doorstep Parcels, Speed Sketch, and Find the Zoo Keeper demos and Python, JavaScript, and REST guides. Ecosystem integrations are rolling out across OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit, and OpenCode alongside the google-genai SDKs, with support coming soon to Gemini in Google Cloud. OpenRouter already lists google/gemini-decision-1. Sundar Pichai posted at 2026-11-11T18:15:20Z; Google for Developers @googledevs followed at 18:26:24Z; Logan Kilpatrick @OfficialLoganK said at 18:16:26Z that you can try it for free in the API.

Key points:

Explain

Why it matters

This is a formal name Google published on November 11, 2026, in an official blog post and in a post from the CEO’s account, not a rumor or an internal nickname. It turns “score a closed set of options instead of writing prose and parsing it” into a nameable model (Gemini Decision 1) and API (Gemini Decisions API), available the same day through the Gemini API in Google AI Studio. For anyone tracking agent routing, policy checks, or low-latency control planes, this is a new name that deserves its own entry.

This page only restates public official materials. Figures such as 150–250 ms latency, $0.10 per 1M input tokens, zero output-token charges, “nearly 5x cheaper,” lowest calibration error, and Gemini Notebook’s “up to 40% lower latency” are the company’s; this page does not independently reproduce them.

Limits

Public materials do not include a full benchmark table, independent third-party reproductions, or a date for Gemini on Google Cloud. A confidence score is a model output, not a guarantee of correctness; the official post itself suggests routing ambiguous cases to a fallback or a reviewer. Do not treat this entry as a production routing or safety guarantee.

Usage

  1. Read the official blog at https://blog.google/innovation-and-ai/technology/developers-tools/gemini-decisions-api/ , and compare it with Sundar Pichai’s announcement at https://x.com/sundarpichai/status/2120581155078729728 , the Google for Developers post at https://x.com/googledevs/status/2120583940247552381 , and Logan Kilpatrick’s note at https://x.com/OfficialLoganK/status/2120581433639198885 ;
  2. Follow the official instructions to get an API key in Google AI Studio and call gemini-decision-1 / the Decisions API; see https://openrouter.ai/google/gemini-decision-1 for the OpenRouter listing, and treat Google Cloud support as forthcoming until official docs say otherwise;
  3. When you cite latency, pricing, the “nearly 5x cheaper” comparison, calibration error, or internal use cases, attribute them to the 2026-11-11 official blog or announcement posts and re-check the upstream source;
  4. Do not describe Gemini Decision 1 / Gemini Decisions API as a general chat model, as the Interactions API, or as Microsoft-Decision-1; this page only restates the official blog and announcement posts, so re-check the upstream source before you cite it elsewhere.

Advertisement