Release2026-10-02
Perplexity released pplx-decider-v1-27b, a decision model fine-tuned from Qwen3.8-27B. It scores 85.71% across 11 benchmarks. The model classifies text and images into user-defined categories. It returns a selected choice and calibrated probabilities. It requires Python 3.12 and a CUDA GPU with about 49 GiB of memory. An inference example script is provided for direct use. The model is not yet deployed by any inference provider. It has been downloaded 165 times in the last month.
What it unlocks
Classifying text and images into predefined categories with calibrated probabilities.
- 26B parameters
- 49 GiB VRAM required
- 85.71% overall benchmark score
- 165 downloads last month
What you need to act on it
- Python 3.12+
- CUDA GPU with ~49 GiB VRAM
Sources