SYLIN

EmbeddingGemma

by Google · Model · Language & code models

Open multimodal embedding model: 270M text, +170M vision, +300M audio; runs on phones; about 1.5 GB, Apache 2.0

AccessOpen weightsNeeds~1.5 GBDownload1.5 GBLicenseApache-2.0

Evidence

  1. Oct 11
    Shown scoring highest among similar-sized embedding models on image and audio embedding benchmarks; context 5.5 min audio, 29 images or 59 frames

Cost to run

Alternatives

Nemotron EmbedRanked first at the 1B scale on the retrieval leaderboard, scoring 72.4% (RTEB), a 27% error reduction over its predecessor
Made from