NV-Embed-v2 is NVIDIA's high-performance embedding model, ranking at the top of MTEB benchmarks. Optimized for retrieval tasks with 4096 token long context support, it's the ideal choice for enterprise RAG and search applications.

Core Features

MTEB #1: Top MTEB leaderboard ranking
Long Context: 4096 tokens support
Retrieval Optimized: Designed for RAG
Fast Inference: GPU-accelerated
Open Source: Model weights available

Performance

MTEB Average: 69.3 score (Rank #1)
Retrieval: Industry-leading nDCG@10
Classification: High accuracy
Semantic Similarity: Precise matching

Use Cases

RAG system document embedding
Enterprise semantic search
Q&A system retrieval
Document similarity computation
Knowledge graph construction

Deployment

NVIDIA API: Cloud API
Local: GPU inference
Optimization: TensorRT acceleration

Summary

NV-Embed-v2, with top MTEB performance, is the best embedding model for retrieval tasks. Long context and open-source nature make it ideal for enterprise RAG applications.

All Tags

Items tagged with High Performance (1)

NV-Embed-v2