The Nemotron 3 Embed model now ranks first overall on the Retrieval Targeted Evaluation Benchmark. It outperforms previous iterations by optimizing dense retrieval for agentic workflows. This improvement reduces noise in RAG pipelines. Developers can now leverage higher precision context retrieval to minimize hallucinations in complex, multi-step AI agents.