Perplexity: Decider Flash V1.2

Perplexity: Decider Flash V1.2 is a decision model from Perplexity, released by the provider on 2026-10-16. It belongs to the decisions category and is positioned as the smaller sibling of Decider V1.2, sharing the same API contract as that larger model. It is a 4B-parameter model, which makes it the lighter option in the Decider line for workloads where the decision task, rather than open-ended text generation, is the point.

The model accepts text and image inputs and produces decisions as its output modality. It does not generate prose; instead, it reads the content passed to it and returns a decision. The context window is 262144 tokens, and the maximum output is 235929 tokens. That combination of a large input budget with a decision-shaped output means you can hand it long documents, transcripts, or mixed text and image material and receive a structured judgment rather than a written answer.

Practical use cases follow directly from that shape. Routing and triage are natural fits: classifying incoming requests, tickets, or messages into categories, or deciding whether an item should be escalated. Moderation and policy checks are another, since the model can read text alongside images and return a decision on the content. Approval-style workflows, such as accepting or rejecting a submission against criteria you supply in the prompt, also map onto a model whose only output is a decision. Any pipeline where a downstream system consumes a verdict rather than reading generated text can use it without parsing free-form prose.

You call it through api14's OpenAI-compatible API using the model id "perplexity/pplx-decider-flash-v1.2". Because the endpoint follows the OpenAI chat completions shape, you can point existing OpenAI SDK clients at api14 and change only the model id. The model lists no supported parameters beyond the standard request fields, so keep your request body to the basics: the model id, your messages, and the content you want a decision on. There are no documented tuning knobs such as temperature or top_p to configure, so behavior is determined by the input you provide.

If your task requires generated text, explanations, or conversational replies, this model is not the right fit, since its output modality is decisions only. For heavier decision workloads, the larger Decider V1.2 is the sibling to consider, as Flash V1.2 is the smaller, lower-cost variant with the same API contract.