Perplexity: Decider Flash Latest

Perplexity: Decider Flash Latest is a model from Perplexity that belongs to the decisions category. Its model id on api14 is ~perplexity/pplx-decider-flash-latest, and the name reflects that it is a moving target: Perplexity describes it as a model that always redirects to the latest model in the Decider Flash family. If you point your integration at this id, you are choosing the family rather than a frozen snapshot, and the underlying model behind the id can change as Perplexity updates the line. The provider release date listed for this entry is 2026-10-16.

The model accepts text and images as input and produces decisions as output. That output modality is the defining characteristic here: this is not a general chat or text generation model, it is built to return a decision rather than prose. The context window is 262144 tokens, which is large enough to hold substantial documents, long conversation histories or multiple images alongside instructions in a single request. The maximum output is listed as 235929 tokens.

Because the inputs are text and images and the output is a decision, the practical uses follow from routing and classification style work. You can send a request that combines written context with one or more images and expect a decision back, which suits workflows where something must be chosen, approved, rejected, routed or flagged rather than explained. Examples include triaging incoming requests, selecting a category or action for a piece of content, or making a call on a case that includes a screenshot or scanned document. The large context window means you can supply lengthy background material without trimming it down first, and the large maximum output gives room for decisions that carry substantial supporting structure.

You call it through api14's OpenAI-compatible API using the model id ~perplexity/pplx-decider-flash-latest, in the same way you would call any other chat completions style endpoint. The listed supported parameters for this model are an empty set, so you should not assume that sampling controls, tool definitions or response format options will be accepted or have an effect. Send your text and image content in the request and read the decision from the response.

There is one practical caveat worth weighing before you build on it. Because the id always redirects to the latest model in the Decider Flash family, behaviour behind the same model id can shift over time, which matters if you need reproducible output for evaluation or regression testing. If your task needs generated text rather than a decision, or if you depend on tuning parameters that this model does not list as supported, a different model is likely a better fit.