AI news ·
Perplexity releases pplx-decider-v1.2 for text and image decisions
Perplexity released a 27-billion-parameter model that evaluates related questions and high-resolution images, with internal benchmarks provided and a recommendation for independent testing.

Perplexity has released pplx-decider-v1.2, a 27-billion-parameter model designed to evaluate multiple related questions in a single sequence and process high-resolution images. The model is positioned for use across several fields, including creative work, education, healthcare, IT and development, product development, sales, science and research, and writing.
The performance benchmarks reported for this release were conducted internally by Perplexity. Teams evaluating the model should independently test how well it fits their specific workloads and examine any effects that correlated decisions may have on output quality.