ModelsMedium importance
Perplexity launches Decider v1.3, open models that answer with probabilities instead of text and now locate exact passages
Perplexity published pplx-decider-v1.3, with 27 billion parameters, built on Qwen3.8-27B, and pplx-decider-flash-v1.3, with 4 billion, built on Qwen3.5-4B, with weights on Hugging Face under the Apache 2.0 license. They are “decision models”: instead of writing an answer, they return the probability of each option in yes-or-no, multiple-choice or score questions about a text or image. The new feature in this version is Locate, which extracts exact text passages by type or instruction. According to Perplexity's own figures, the larger model averages 91.6% accuracy on 63 query-classification tasks; Locate is the weakest point, with an average of 69. The models are also available in Perplexity's Decisions API and on OpenRouter.
Published in edition no. 28 of Noticiário do Joãozinho ·
Translated by Joãozinho AI from the original Brazilian Portuguese edition. Facts, numbers and sources are the same as in the original; quotes, video titles and source names stay in their original language. Read the original in Portuguese
Original source
MarkTechPostPerplexity: Decider v1.3
https://www.marktechpost.com/2026/11/13/perplexity-decider-v1-3-open-weight-decision-models-locate/