Google has introduced EmbeddingGemma 2, an open embedding model featuring 740 million parameters. Despite its relatively compact size, the model reportedly outperforms some competitors that are twice as large.
According to The Decoder, Google claims that EmbeddingGemma 2 achieves superior performance while requiring only 191 MB of RAM, highlighting its efficiency in resource usage compared to larger models.
This development may hold particular interest for Japanese technology sectors focused on AI and machine learning, where efficient models can support faster deployment in both FX and equity trading platforms, as well as crypto analytics.