Google's EmbeddingGemma 2 outperforms rivals twice its size
google released embeddinggemma 2, a 740m parameter model that outperforms rival embedding models up to twice its size on multimodal embedding benchmarks, and it can run locally without an api key.
google released embeddinggemma 2, an open model with 740 million parameters that converts text, images, video, audio, and code into numerical vectors. it achieves a score of 78.68 on the massive text embedding benchmark (code), surpassing its predecessor's score of 68.76. embeddinggemma 2 can run locally without an api key, processing queries in 20-70 milliseconds via webgpu in the browser, and requires only around 191 mb of ram. paired with small open models like gemma 4, it enables offline rag apps without sending data to external servers.