{"type":"post","id":"google-expands-embeddinggemma-beyond-text-to-images-audio-3704a7","slug":"google-expands-embeddinggemma-beyond-text-to-images-audio-3704a7","title":"Google expands EmbeddingGemma beyond text to images, audio and video","description":"**Thesis:** Multimodal embeddings become more usable when one compact representation layer can index text, images, audio and video together, including on local devices.","retrieval_nugget":"**Thesis:** Multimodal embeddings become more usable when one compact representation layer can index text, images, audio and video together, including on local devices.","published_at":"2026-10-10","updated_at":"2026-10-10","record_date":"2026-10-10","date_kind":"published_at","topics":["google"],"source_urls":["https://siliconangle.com/2026/10/06/google-expands-embeddinggemma-beyond-text-to-images-audio-and-video"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:google-expands-embeddinggemma-beyond-text-to-images-audio-3704a7","status":"published","visuals":[],"editorial_provenance":{"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:93db2cf3c8262de7be8f50ff8409d8e27f91ffd00bc682aeebefd410e9f1a01d","reviewed_at":"2026-10-10T09:04:10.120Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":3,"watch_condition_count":1,"related_record_count":0},"analysis":{"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"Multimodal embeddings become more usable when one compact representation layer can index text, images, audio and video together, including on local devices.","observed_facts":[{"text":"SiliconANGLE reports that EmbeddingGemma 2 has 740 million parameters and its 270-million-parameter text core used about 191 MB in a quantized Pixel 11 Pro test.","source_urls":["https://siliconangle.com/2026/10/06/google-expands-embeddinggemma-beyond-text-to-images-audio-and-video"]},{"text":"The report says the model maps text, images, audio and video into one embedding space.","source_urls":["https://siliconangle.com/2026/10/06/google-expands-embeddinggemma-beyond-text-to-images-audio-and-video"]},{"text":"The report says 768-value vectors can be reduced to 256 while retaining about 95% of image, video and speech retrieval quality, according to Google.","source_urls":["https://siliconangle.com/2026/10/06/google-expands-embeddinggemma-beyond-text-to-images-audio-and-video"]}],"mechanism":"A shared embedding space lets an application compare different media types through a common retrieval layer. Local execution can keep that retrieval near the data, while dimensionality reduction lowers the storage cost of indexing media.","why_now":"The model weights are available under Apache 2.0 on Hugging Face and Kaggle, and the source says they work with vLLM, llama.cpp and Ollama. That makes the release actionable for developers evaluating local multimodal retrieval rather than a future-only product promise.","implications":["Teams can test workflows such as locating a video moment from a spoken or typed query without sending the media off-device.","The 256-dimensional result suggests a practical storage-quality tradeoff, but teams still need to validate their own corpora and retrieval tasks.","An embedding layer alone does not establish end-to-end application quality; indexing, evaluation and interface work remain necessary."],"evidence_boundary":"The measurements and quality-retention figure are reported by SiliconANGLE from Google's claims and test context; the supplied source does not provide independent replication or comparative results across production workloads.","watch_conditions":["Whether independent evaluations confirm retrieval quality and memory use across real multilingual, noisy and long-form media collections will determine how broadly the on-device claim transfers."],"related_records":[],"new_branch_reason":"No prior New Runtime record is supplied, so this release starts a distinct analytical record focused on multimodal embedding deployment and local retrieval."},"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/google-expands-embeddinggemma-beyond-text-to-images-audio-3704a7/","markdown":"https://newruntime.com/posts/google-expands-embeddinggemma-beyond-text-to-images-audio-3704a7.md","json":"https://newruntime.com/posts/google-expands-embeddinggemma-beyond-text-to-images-audio-3704a7.json"},"next_reads":[{"type":"topic","path":"/topics/google/","url":"https://newruntime.com/topics/google/","title":"Google","reason":"Explore the google topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/google.md","json_url":"https://newruntime.com/topics/google.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/source-ledger.json","url":"https://newruntime.com/source-ledger.json","title":"Source Ledger","reason":"Inspect the public source evidence graph.","media_type":"application/json"}]}