AIThe New Stackabout 2 hours ago
Google Releases Compact EmbeddingGemma 2
4 min read
Google released EmbeddingGemma 2, a 740-million-parameter model running on a Pixel 11 Pro using 567MB of active RAM. The company quadrupled the context window to 8,192 tokens and introduced Matryoshka Representation Learning to shrink the index size. Weights are available on Hugging Face and Kaggle.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyGoogle
