✦
Summary
NVIDIA Nemotron 3 Embed is a collection of embedding models for retrieval, RAG, code search, agent memory, and agentic retrieval systems. The release includes an 8B flagship quality model and efficient 1B variants, giving developers options across retrieval accuracy, memory footprint, throughput, and deployment hardware.
Key Features
Provides open embedding models for RAG, agentic retrieval, code retrieval, and memory systems.
Includes an 8B BF16 flagship model for maximum retrieval quality.
Includes 1B BF16 and 1B NVFP4 variants for efficient production deployment.
Ranks the 8B model first on RTEB in the cited release.
Supports 32k context windows for long-input retrieval.
Uses bidirectional adaptation of instruction-model backbones for embedding generation.
Compresses smaller variants through structured pruning and teacher distillation.
Offers an optimized NVIDIA NIM serving path for production-scale retrieval systems.
About NVIDIA Nemotron 3 Embed
The 8B model adapts a Ministral-3-8B-Instruct backbone into a bidirectional encoder for full-sequence retrieval, while the 1B models are produced through structured pruning and distillation from larger teacher checkpoints. The collection includes BF16 variants plus an NVFP4 Blackwell-optimized model that quantizes weights and activations for high-throughput serving.
Nemotron 3 Embed is useful for teams building production retrieval pipelines where better embeddings reduce repeated searches, irrelevant context, and downstream agent token cost. NVIDIA reports the 8B model ranks first on RTEB, strong MMTEB and LongEmbed results, and optimized NIM serving support for enterprise retrieval workloads.