Perplexityapipplx-decider v1.3
New open-weight Perplexity models that answer yes/no, multiple-choice, and scored questions about text, now with a feature that pulls exact passages out of the input.
Details
- Two models: Decider v1.3, the larger one built on Qwen3.8-27B for nuanced judgments and detailed rubrics, and Decider Flash v1.3, a smaller distilled version for high-volume work where speed and cost matter most.
- Each question returns a probability for each answer, and several questions about the same input run in one pass.
- New in v1.3 is locate, which returns the exact matching text with its start and end character positions, or an empty list if nothing matches. It targets finding names, spotting personal information, and answering questions by extracting text.
- Both models are downloadable from Hugging Face under an Apache license and also available through OpenRouter and the Perplexity Decisions API. Perplexity reports about $26.47 per million production requests for v1.3 and $13.24 for Flash on its own workload.
Best for
High-volume labeling, routing, and filtering, plus pulling names or personal details out of text.
When to use it
You need a low-cost classifier for thousands of items instead of calling a general chat model each time.
Who benefits
Developers and data teams building on the Perplexity Decisions API.
Join the newsletter to get ahead with AI.
I share strategies, new modules, free guides, and tools to get you better at using AI in your work.
Leading a team? See team programs →