Gemini News 10/06/2026 Read AI Rating: Medium
Google releases EmbeddingGemma 2, an open multimodal embedding model for text, images, audio and video
EmbeddingGemma 2 has 740 million parameters, an 8K-token context window and an Apache 2.0 license, and maps text, images, audio and video into one embedding space. Quantized text-only weights need about 191MB of RAM on a Pixel 11 Pro.