ZeroEntropy's reranker family includes zerank-1, zerank-1-small, and zerank-2 — state-of-the-art neural models designed to improve retrieval accuracy in production AI systems. These models take candidate results from an initial search or vector lookup and rescore them using deeper semantic understanding, surfacing the most relevant chunks for downstream language models and agents.

The rerankers are available via API and through Hugging Face, with benchmarks showing industry-leading accuracy on standard information-retrieval evaluations.