Snowflake

SwiftKV reduces prefill compute by up to 50% by combining model rewiring and knowledge-preserving self-distillation.

A collection of text embedding models optimized for retrieval accuracy and efficiency