Is Gemini Decision 1 state of the art?Pending benchmarks.Announced Nov 11, 2026; no independent results yet.
Google DeepMind's multimodal decision model for the Gemini Decisions API. It scores up to 1,000 independent closed-set questions in one pass over a shared text, image, audio, video, or PDF input of up to 262,144 tokens, and returns calibrated probabilities and confidence scores with zero output tokens. Announced November 11, 2026. Available through the Gemini API in Google AI Studio as gemini-decision-1. No public weight release is stated.
The details so far
- Lab
- Google DeepMind
- Announced
- Nov 11, 2026
- API
- Live · gemini-decision-1
- Open weights
- Closed · No public weight release stated on the official blog or docs
- Type
- Specialized multimodal model for the Decisions API. It evaluates typed questions over a shared input in a single forward pass and returns calibrated probabilities, with no generated answer text (Google blog and docs)
- Question types
- Yes/no (noul), returning a probability from 0.01 to 0.99; choice, selecting from 2 to 255 options; and score, grading against 2 to 255 ordered rubric levels. Every answer includes a confidence score from 0 to 1 (Google blog and docs)
- Context
- Shared input up to 262,144 tokens (256K tokens on the Google blog). A request can include 1 to 1,000 independent questions; the shared input is billed once (Google docs and blog)
- Modalities
- Text, image, audio, video, and PDF input. Output is calibrated probabilities only (zero output tokens) (Google blog and docs)
- Hosted access
- Live in the Gemini API and Google AI Studio via client.decisions.create on API version v1beta, model id gemini-decision-1. Requires the google-genai SDK 3.2 or later. Official docs say ecosystem integrations are rolling out, with Gemini in Google Cloud coming soon (Google blog and docs)
- Pricing
- Paid tier: $0.10 per 1M input tokens (text, image, video, audio, PDF). Output is not billed (0 output tokens). Free-tier input is free of charge in Google AI Studio, with free-tier prompts used to improve Google products. Context caching and the Batch API are not available (Google blog and Gemini Developer API pricing)
- Latency
- Official blog: 150-250 ms end-to-end, including 50-100 ms of model compute, for multimodal inputs and contexts up to 16K tokens, remaining nearly flat from 1 to 100 questions per request
- License
- Not stated as a weight license. Access is under the Gemini API / Google AI Studio terms
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: self-reported as leading a lab benchmark suite covering agent routing, reasoning, ranking, and multimodal understanding, with the lowest calibration error on that suite (Google blog).
Lab claim: self-reported as nearly 5 times cheaper across the shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API, with zero output tokens billed (Google blog).
Sources
LABGemini Decisions API: Fast, calibrated multimodal decisions (Google blog)blog.google/innovation-and-ai/technology/developers-tools/gemini-decisions-api/ · Nov 11, 2026LABGemini Decision 1 model card (Gemini API docs)aistudio.google.com/docs/models/gemini-decision-1LABDecisions API (Gemini API docs)aistudio.google.com/docs/decisions-apiLABGemini Developer API pricingaistudio.google.com/docs/pricingLABLogan Kilpatrick on X: Gemini Decision 1x.com/OfficialLoganK/status/2120581433639198885 · Nov 11, 2026Tracking continues until an independent board lists it.