Home/Events/NVIDIA Nemotron 3.5 Embed, Rerank, and Parse: Advancing Agents with World-Class Retrieval Models

NVIDIA Nemotron 3.5 Embed, Rerank, and Parse: Advancing Agents with World-Class Retrieval Models

Confirmed

Confidence

90%

Impact: 80%

Updated 1h ago

Consensus Brief

NVIDIA has expanded its Nemotron Retriever collection with new models for embedding, reranking, and document extraction, specifically the Nemotron 3.5 series. These models enhance retrieval quality and efficiency for enterprise agents, allowing them to produce grounded answers with reduced inference costs.

Sourced from

Hugging FaceT1

Primary: Hugging Face

What Changed Since Last Update

1h ago

The introduction of Nemotron 3.5 models, including Embed and Rerank variants, improves multimodal retrieval and relevance ranking capabilities compared to previous versions.

Claim Ledger

4 claims tracked across sources

Confirmed Fact

Nemotron 3.5 Embed 8B BF16 achieves 61.17 nDCG@10, ranking first among dense retrieval models on ViDoRe v3.1.

Confirmed Fact

Adding Nemotron 3.5 Rerank reduces agentic token cost per query by approximately 14% with Nemotron 3.5 Embed 8B.

Official Claim

The models support multilingual retrieval across 34 languages.

Confirmed Fact

Nemotron embedding fine-tuning recipes support custom document collections, achieving relative nDCG@10 gains of 18% and 3%.

Role-Based Impact Analysis

Source Timeline

1 source corroborating

NVIDIA Nemotron 3.5 Embed, Rerank, and Parse: Advancing Agents with World-Class Retrieval Models

Hugging Face·1h ago