Perplexity releases open-weight pplx-decider-v1.2 models, ranked #2 on Decision Index

Perplexity releases pplx-decider-v1.2, a 27B open-weight multimodal model, and pplx-decider-flash-v1.2, a 4B version, on Hugging Face.

Perplexity says pplx-decider-v1.2 ranks second overall on the Hugging Face Decision Index and Flash is the top 4B model on the index. Both are also available through the Perplexity Decisions API, where Perplexity prices pplx-decider-v1.2 at $0.02 per million input tokens and Flash at half that. Perplexity says text and vision scores improve over the prior generation and inference runs up to 5x faster because all questions in a request run in one pass. Perplexity positions pplx-decider-v1.2 for nuanced decisions and Flash for high-volume tasks with clear criteria, such as routing, labeling, and filtering. The ranking and performance figures come from Perplexity's own posts.