Perplexity posts pplx-decider-v1.2, an open-weight decision model for text and images
Perplexity's Hugging Face card for pplx-decider-v1.2, updated Oct 16, describes the next version of its pplx-decider-v1.1-27b decision model. It is fine-tuned from Qwen3.8-27B and listed under the Apache 2.0 license. Instead of writing text, it returns probabilities for multiple-choice, true/false and score questions. It now reads a whole request in one sequence, so several questions about the same input cost about as much as one, and it accepts images up to 1.6 MP. The card says it needs a CUDA GPU with room for about 52 GiB of BF16 weights.
How you’d use it: Route support tickets to the right team with one pplx-decider request.