Gemini Decision 1 and Decider v1.3: A Decision Layer in Front of Your Agent
Quick summary: Google released the Gemini Decisions API on November 11, and Perplexity followed with open weights for Decider v1.3 the next day. Both separate decisions over fixed options from text generation, making them useful for routing, checks, and filtering before an agent runs. What matters is more than skipping a written response: it is a single forward pass, question isolation, and…
Google unveiled the Gemini Decisions API on November 11, followed by Perplexity's launch of Decider v1.3 and Flash v1.3 the following day. These separate decision-making tools aim to filter options before sending queries to an agent. Instead of generating text, they return structured outputs like probabilities and fixed options. This article examines how they function, the tradeoffs between cloud-hosted and self-hosted deployments, and a code snippet that prioritizes decisions before invoking a generative model.
One key advantage of decision APIs is separating the question of "should we proceed" from "how should we respond." Google's Decisions API focuses on tool selection, policy checks, model routing, and context scoring, while Perplexity's Decider v1.3 and Flash v1.3 serve a similar purpose. By handling decisions as discrete tasks, developers can streamline business logic and avoid over-relying on generative models for simple filtering tasks.
Google emphasizes that decision APIs return typed answers directly, including probabilities and fixed options, rather than generating text. This approach reduces latency and eliminates the need to parse and interpret text tokens, which can be brittle and slow. With end-to-end latency reported at 150-250 milliseconds, decision APIs offer faster responses compared to traditional generative models. However, they still require careful calibration of thresholds and handling of errors and abstentions.
While decision APIs streamline decision-making, they do not replace the need for natural language generation in all scenarios. For tasks requiring free-form explanations or dynamic category discovery, enumerating fixed outcomes may not be appropriate. Both Google and Perplexity provide APIs for decision-making, with Apache-2.0 open weights available for deployment.
However, pricing varies: Google charges $0.10 per million input tokens, while Perplexity's costs are not specified. It's essential to consider input token usage and the potential costs of subsequent generative model calls when using decision APIs.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
