According to CoinMeta, Perplexity has open-sourced two multimodal decision-making models, namely pplx-decider-v1.2 (27B) and pplx-decider-flash-v1.2 (4B), which are based on qwen3.8-27b and qwen3.5-4b. These models are not designed to generate long answers but rather to perform classification, scoring, and judgment tasks based on text or images. They can be used in scenarios such as agent task routing and content moderation. Compared to the previous generation that inferred on a question-by-question basis, the new version allows multiple questions to share the same input, enabling judgments to be made in a single inference process. Perplexity claims that in some scenarios, the inference speed has increased by up to five times. Both models use the Apache 2.0 license and integrate with Perplexity API and Vercel AI Gateway. API charges $0.02 per million inputs for token and $0.01 per million inputs for the other model, while the output is provided for free.