Skip to main content
Modelembeddingsv2.0

NV-Embed-v2

by NVIDIA · free · Last verified 2026-03-17

NV-Embed-v2 is NVIDIA's second-generation text embedding model, achieving top-1 performance on the MTEB benchmark at release with innovative decoder-only LLM architecture adaptations for embedding tasks. It introduces latent attention layers and removes causal masking to produce richer, task-aware text representations.

https://huggingface.co/nvidia/NV-Embed-v2
C+
C+Average
Adoption: C+Quality: A+Freshness: ACitations: BEngagement: F

Specifications

License
NVIDIA Open Model License
Pricing
free
Capabilities
text-embeddings, semantic-search, retrieval, reranking, classification
Integrations
NVIDIA NIM, Hugging Face, LlamaIndex, LangChain
Use Cases
rag-pipelines, enterprise-search, semantic-retrieval, document-ranking
API Available
Yes
Parameters
7.8B
Context Window
32K
Modalities
text
Training Cutoff
2024
Tags
nvidia, embeddings, open-source, retrieval, state-of-the-art
Added
2026-03-17
Completeness
100%

Index Score

56.9
Adoption
58
Quality
91
Freshness
86
Citations
62
Engagement
0

Explore the full AI ecosystem on Agents as a Service