Uploaded November 2026 | Updated November 2026, 22 hours ago
Introducing Microsoft Embedding 1, a Multimodal Embedding Model
Most of the answers in your enterprise documents aren't written as sentences. They're in a chart on page 14, a row in a pricing table, or a diagram in a scanned manual. Text search can't see them. Microsoft Embedding 1 can.
What you get:
• Retrieval that reads the page the way people do: text, tables, charts, and layout together
• The highest published ViDoRe v3 score for a single retrieval model, the benchmark built on real enterprise documents
• Multi-vector output for dense visual pages, with control over vector count and size to fit your storage budget
• A faster path to RAG and agents that answer from your contracts, reports, and decks
Now in preview in Azure Cosmos DB, coming soon to Microsoft Foundry.
Try it: aka.ms/MicrosoftEmbeddingOne
#MicrosoftEmbedding1 #EnterpriseSearch #RAG
Introducing Microsoft Embedding 1, a Multimodal Embedding Model
Most of the answers in your enterprise documents aren't written as sentences. They're in a chart on page 14, a row in a pricing table, or a diagram in a scanned manual. Text search can't see them. Microsoft Embedding 1 can.
What you get:
• Retrieval that reads the page the way people do: text, tables, charts, and layout together
• The highest published ViDoRe v3 score for a single retrieval model, the benchmark built on real enterprise documents
• Multi-vector output for dense visual pages, with control over vector count and size to fit your storage budget
• A faster path to RAG and agents that answer from your contracts, reports, and decks
Now in preview in Azure Cosmos DB, coming soon to Microsoft Foundry.
Try it: aka.ms/MicrosoftEmbeddingOne
#MicrosoftEmbedding1 #EnterpriseSearch #RAG










