Perplexity
Models published by Perplexity, available through the AnyRouter API. Each can route across multiple upstream providers for availability and price.
PPLX Decider Flash v1.3 (4B) is the small decision model of Perplexity's v1.3 family, distilled from the 27B. It returns a probability distribution over fixed answers, not chat text. State can be text or images, and it supports Locate.
PPLX Decider v1.3 (27B) is Perplexity's current multimodal decision model. It returns a probability distribution over a fixed set of answers, not chat text. State can be text or images, and it adds Locate span extraction.
Sonar Pro is Perplexity's multimodal chat model with a 200,000-token context window. It accepts text and image input and streams text output.
Sonar is Perplexity's lightweight, fast chat model with citations and customizable sources. It accepts text and image input and has a 127,000-token context window.
PPLX Decider Flash v1.2 is Perplexity's on-device decision model, run by the Decider for Mac app on your own paired Mac. It returns a probability distribution over a fixed set of answers to typed questions, not chat text. Only your own Mac serves your requests.
pplx-embed v1 4B is Perplexity's larger text embedding model for web-scale dense retrieval. It returns 2560-dimension vectors and accepts 32K-token input.
pplx-embed v1 0.6B is Perplexity's lightweight text embedding model for web-scale dense retrieval. It is aimed at low-latency use and accepts 32K-token input.
PPLX Decider v1.1 (27B) is a newer checkpoint of Perplexity's multimodal decision model, with the same contract as v1. It returns a probability distribution over a fixed set of answers, not chat text. State can be text or images, with 262,144 tokens of context.
PPLX Decider v1 (27B) is Perplexity's multimodal decision model with 27B parameters. It returns a probability distribution over a fixed set of answers, not generated chat text. State can be text or images, and the context window is 262,144 tokens.