Google DeepMind Releases EmbeddingGemma 2, a 740M Open Multimodal Embedding Model Built on Gemma 4

modelsMarkTechPostOct 6, 2026

Google DeepMind's EmbeddingGemma 2 maps 5 input types into one 768d space and ships today under Apache 2.0. The post Google DeepMind Releases EmbeddingGemma 2, a 740M Open Multimodal Embedding Model Built on Gemma 4 appeared first on MarkTechPost .

Impact: MediumConfidence: LowNovelty: Incremental
modelsmarktechpostmultimodal

Read the original on MarkTechPost

See the whole story

Sens.ai aggregates and assesses published reporting. The assessment above is machine generated; the original source is authoritative.