AI Research Atlas

Gemini Embedding 2

Google DeepMind · 10 March 2026

Gemini Embedding 2: Google's first multimodal embedding model maps text, images, video, audio and PDFs into one embedding space.

Unified vector space lets one index serve cross-modal search and RAG. GA 2026-04-22; File Search added multimodal image search with it on 2026-05-05.

Date
Tuesday, 10 March 2026
Lab
Google DeepMind
Kind
model
Access
closed API

Sources

  1. ai.google.dev/gemini-api/docs/changelog
  2. ai.google.dev/gemini-api/docs/embeddings

This record was checked against its sources on 6 October 2026. How we check

Related