Perplexity's Decider V1.2 models now available on AI Gateway
Perplexity’s Decider V1.2 models, including a 27B and a 4B variant, are now accessible via AI Gateway for decision-making tasks such as routing, scoring, and classification.
Perplexity’s Decider V1.2 27B and Decider Flash V1.2 4B models are now available on AI Gateway, offering multimodal decision capabilities for classifying content, routing requests, and scoring text or images. The models process typed questions, returning probabilities for Boolean queries, selecting categories for Choice questions, and evaluating content against rubrics for Score questions. Applications can use these outputs to automate actions like routing support tickets or flagging issues for human review.
Decider V1.2 27B is designed for nuanced judgments and detailed rubrics, while Decider Flash V1.2 4B is optimized for high-volume labeling, filtering, and routing with clear criteria. Both models support multiple questions sharing the same input in a single request, enabling efficient batch processing. Image inputs are handled by embedding base64-encoded image_url objects in JSON via the AI SDK or TypeSafe-compatible API.
The models can be tested in Perplexity’s model playground or accessed through AI Gateway, which provides a unified API key for calling models and monitoring usage and costs. Decision requests are logged and included in spend budgets, with inference pricing based on provider rates without additional markup or platform fees.
AI Gateway aggregates decision models, allowing users to browse available options and integrate them into applications. The service supports both Perplexity-hosted models and user-supplied provider credentials through Bring Your Own Key (BYOK), ensuring flexibility in deployment and cost management.