MongoDB's research division has introduced a new knowledge distillation framework for text embedding models called LEAF (Lightweight Embedding Alignment Framework). Alongside the framework, the company released two new open-source embedding models trained with LEAF. These models are designed to make advanced AI models smaller, faster, and more flexible for various applications. The models, named mdbr-leaf-ir and mdbr-leaf-mt, are tailored for tasks such as semantic search, retrieval-augmented generation (RAG), classification, and clustering. A key feature is their suitability for CPU-only deployments, as well as for IoT and mobile devices, which expands their potential use cases. This release is presented as a contribution to the open-source community under a permissive Apache 2.0 license.