Perplexity's Decider V1.2 Models Arrive on Vercel's AI Gateway
Vercel has made two Perplexity decision models, Decider V1.2 27B and Decider Flash V1.2 4B, available on its AI Gateway. Both are multimodal models built for classifying content, routing requests, and scoring text, structured records, and images.
- Perplexity Decider V1.2 27B and Decider Flash V1.2 4B are now available on Vercel's AI Gateway.
- Both multimodal models answer Boolean, Choice and Score questions with probabilities.
- The 27B model targets nuanced judgments and detailed rubrics; the 4B Flash model targets high-volume labeling, filtering and routing.
- Access is available through the AI SDK, the Decisions API and the TypeSafe-compatible API.
- Vercel says inference uses provider pricing with no markup or platform fee, including with BYOK.
What the models do
According to Vercel, the models answer typed questions with probabilities rather than free-form text. Three question types are supported:
- Boolean: returns the likelihood of yes.
- Choice: selects from categories that the developer defines.
- Score: assesses content against an ordered rubric.
Vercel says applications can use these results to route a support ticket, flag a damaged product, or send an uncertain decision for human review. Multiple questions can share the same input in a single request.
Two sizes for different workloads
Decider V1.2 27B, with the model ID perplexity/pplx-decider-v1.2, is the larger of the two. Vercel positions it for nuanced judgments and detailed rubrics.
Decider Flash V1.2 4B, with the ID perplexity/pplx-decider-flash-v1.2, is the smaller model. It is aimed at high-volume labeling, filtering, and routing where the criteria are clear.
Both models can be tried in the AI Gateway model playground, and Vercel points to a decision model catalog for browsing the available options.
How developers call them
Vercel shows the same delivery-report check running through three interfaces: the AI SDK, the Decisions API, and the TypeSafe-compatible API. In the AI SDK example, a developer imports experimental_decide from the ai package and passes the model ID, a state string, and a set of questions. The sample state reads "The tablet arrived with a shattered display and will not turn on." A single Boolean question, whether the text reports a damaged delivery, is defined, and the result is read from result.answers.damaged.
For image input, developers embed base64 image_url objects in JSON state through the AI SDK or the TypeSafe-compatible API. Vercel refers readers to Perplexity's image format and the decision models documentation for request formats and question types.
Pricing and monitoring
AI Gateway lets developers call models with one API key and monitor usage and cost in one place. Vercel states that decision requests are recorded in request logs and included in spend budgets. Inference uses provider pricing with no markup or platform fee, and this also applies when developers supply their own provider credentials through BYOK.
For teams already routing model traffic through AI Gateway, the addition means classification and routing tasks can be handled by dedicated decision models under the same key, logging and budget controls, according to the announcement. The source does not give benchmark results or specific prices for the models.
