Perplexity Releases Decider v1.3: Open-Weight Decision Models Add Locate

What it does

Perplexity has launched Decider v1.3 and Flash v1.3, two open-weight decision models designed to output probabilities instead of natural language text. The latest update adds a feature called Locate, which focuses on identifying precise spans of text relevant to a query. This extension allows the models to perform extraction tasks rather than just classification or ranking.

Why it matters

Unlike typical language models that generate text, Decider v1.3 and Flash v1.3 act as decision engines, providing probabilistic judgments. This shift means developers can embed these models into systems that require more transparent and quantifiable decision-making, improving control over outputs. Adding Locate expands use cases toward tasks requiring exact text snippets, like highlighting, data extraction, or structured question answering. Reporting 91.6% mean accuracy on 63 query-classification tasks signals reliable performance in classification domains, making these models worth considering for workflows demanding probabilistic confidence and span extraction.

Who it is for

Builders and engineers working with natural language processing can leverage these models where prediction confidence and explicit span extraction matter. The open-weight design fits those who want adaptable models without vendor lock-in or reliance on closed APIs. Product teams working on search, question answering, and extraction features can integrate Decider v1.3 to strengthen backend decision components or to augment existing LLM pipelines with more granular, probability-based insights.

The catch

Open-weight models require operators to manage hosting and inference, which adds overhead compared to cloud-hosted APIs. Also, while 91.6% accuracy is solid, it means error rates persist. Proper error handling and fallback strategies remain necessary when replacing or augmenting text generative models with these probabilistic decision tools.

What to watch next

Monitor how Perplexity’s open-weight decision models affect integration strategies in search and extraction-heavy applications. Whether this new Locate capability drives adoption will depend on its ease of integration and actual impact on accuracy for real-world inputs. Also watch if these models are adapted into decision logic workflows beyond NLP, potentially influencing automation and AI-assisted decision systems across sectors.

AI Quick Briefs Editorial Desk