EmbeddingGemma 2
Google's EmbeddingGemma 2 ships with 768-dim and 1024-dim variants, open-weight and free-to-use—a direct move into the embedding-model market dominated by Cohere and Voyage.
Why it matters
Google releases a new open-weight embedding model family, expanding its frontier capability stack and offering practitioners a free alternative to proprietary embedding APIs. The move signals Google's push to compete across the full AI stack, from reasoning models to retrieval infrastructure.
The key facts
7 to knowEmbeddingGemma 2 released by Google
Open-weight model available
768-dimensional and 1024-dimensional variants
Free-to-use offering
Competes with Cohere and Voyage embedding models
Part of Gemma model family ecosystem
Source: simonwillison.net (tech newsletter covering AI releases)
The story so far
Earlier coverage of this storyline
- Google DeepMind Releases EmbeddingGemma 2, a 740M Open Multimodal Embedding Model Built on Gemma 4MarkTechPost
- EmbeddingGemma 2: an open, lightweight multimodal embedding modelGoogle DeepMind Blog
- This story
Go to the source
Simon Willisonsimonwillison.net