Decider Flash V1.2

Perplexity Decider Flash V1.2 is a lightweight decision model that reads text and images and returns probabilities or scores for predefined questions, with applications such as request routing and retrieval filtering. It shares the larger Decider's request format while prioritizing latency and deployment cost, and is designed for structured decisions rather than free-form conversation. The developer's model card says it supports joint questions, but separate requests performed better on many evaluations, unlike the larger model. Weight-based deployment is documented, so teams should check download access and the dedicated inference setup, then compare processing time with the cost of misclassification. Output probabilities alone should not determine automatic approval; representative business cases should establish error patterns and conditions for human review.