opensearch-project / opensearch-neural-sparse-encoding-doc-v3-distill

Overall, the v3 series of models have better search relevance, efficiency and inference speed than the v1 and v2 series. The specific advantages and disadvantages may vary across different datasets.