According to CoinMeta, Perplexity has launched an open-source decision-making model called pplx-decider-v1-27b and released the corresponding decision-making framework API. This model does not generate plain text; instead, it directly returns a probability distribution based on the answers provided by developers. It can be used for classification, request routing, and selection of next steps in agent. The cost of using this model is $0.04 per million inputs of token, while the output is provided for free. Based on the 11 evaluations of 7,210 samples published by Perplexity, the overall accuracy of pplx-decider-v1-27b is 85.71%, which is higher than 84.51% for Jev and 74.76% for qwen3.8-27b. However, Jev scored higher in 6 out of the 11 tests, while Perplexity led in financialphrasebank, ragtruth, tabfact, and circa. This model is specifically designed to handle high-frequency "minor decisions" within agent, such as the allocation of customer service tickets, tool calls, and process selections. Similar models have also been recently introduced, including AWS and Cloudflare.