Gemini Decision 1
Google DeepMind · Released 11 Nov 2026
The model behind the Gemini Decisions API: a new endpoint for fast, closed-set, multimodal decisions. It can evaluate up to 1,000 independent yes/no, multiple-choice, or rubric-scoring questions in one forward pass across a shared input, returning probabilities and confidence scores rather than generated answer text.
- Maker
- Google DeepMind
- Release date
- 11 Nov 2026
- First seen
- 11 Nov 2026
- Access
- Gemini API in Google AI Studio, model id gemini-decision-1, with a no-cost prototyping tier. Google said ecosystem integrations were rolling out across OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit, and OpenCode, with support coming soon to Gemini in Google Cloud.
- Pricing (as publicly stated)
- As stated on 11 Nov 2026: $0.10 per million input tokens and $0 for output tokens. The shared input is billed once per request; each question adds only its own instruction and criteria tokens.
- Modalities
- text, images, audio, video, PDFs
- Question types
- noul, choice, score
- Latency (as stated)
- Google said Gemini Decision 1 returns probabilities in 150-250 ms end-to-end, with 50-100 ms of model compute, remaining nearly flat from 1 to 100 questions per request on contexts up to 16K tokens. Maximum context was stated as 256K tokens.
- Official announcement
- https://blog.google/innovation-and-ai/technology/developers-tools/gemini-decisions-api/
Benchmark claims
These are claims as published by the maker or in the linked official post. We have not independently rerun every benchmark.
Google said that on benchmarks spanning agent routing, reasoning, ranking, and multimodal understanding, Gemini Decision 1 leads across the suite, with the lowest calibration error, and that with zero output tokens billed it is nearly 5x cheaper across the shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API.
Names and marks belong to their owners. This page is an independent summary, not a product page from Google DeepMind.