← Back to live feed1 story
Google has released EmbeddingGemma 2, a 740M parameter AI model that enables natural language or sketch-based searches through video and audio files on local hardware. The system uses Matryoshka embeddings to reduce vector storage to 1/3 while maintaining approximately 95% of retrieval quality. The model supports cross-modal retrieval and is currently the top trending model on Hugging Face.
Content history (2)
- 2026-10-13 · Summary · vi · Replaced stored textGoogle đã ra mắt EmbeddingGemma 2, một mô hình embedding dùng cho tìm kiếm đa dạng dạng t…Người dùng hiện có thể tìm kiếm qua các tệp video và âm thanh bằng ngôn ngữ tự nhiên hoặc…
- 2026-10-13 · Title · vi · Replaced stored textGoogle Launches EmbeddingGemma 2 for On-Device Multimodal SearchGoogle ra mắt EmbeddingGemma 2 phục vụ tìm kiếm đa phương thức trên thiết bị
Sign in to suggest
Key sources
- SOURCE@miaai_lab“Google's EmbeddingGemma 2 is the #1 trending model on Hugging Face right now”x.com
- SUPPORT@theturingpost“1/3 of the vector storage, with ~95% of image, video, and speech retrieval quality, according to Google”x.com
- SOURCEmarketbrief.now