Perplexity
Perplexity

Perplexity Decider v1 27B

Release Date 2026-10-01262K IN · 236K OUTOpen Model(Apache 2.0)
Decision overall3.0Top 46%KO·JA judging3.0Top 44%
AIB DecisionBench-GameResearching
AIB DecisionBench-AgentResearching

Perplexity Decider v1 is Perplexity's model for predefined judgments over text and images, with this entry covering the initial public release. Unlike Sonar, which writes answers from search results, it returns probabilities for classification, rubric-based scoring or whether a supplied condition is met. Adapted from Alibaba's Qwen, it can use documents and images as shared evidence, tying questions and choices to the source rather than generating an open-ended discussion. Published weights provide a self-managed deployment option alongside API access, with outputs organized around specified choices and rating formats. Consult the model detail page's detailed specifications and evaluation analysis for the exact identifier, image and long-input conditions, runtime and billing structure, classification by language and probability calibration.

API cost

$0.02IN$0OUT

Expert Performance Evaluation

Can it understand Korean and Japanese well and classify accurately?AIB DecisionBench-KJTop 45%#12
77.2

More about AIB DecisionBench-KJ

See each evaluation’s detail page for its source and last updateData corrections & removal requests