{"rewrite":{"id":"r_6e8ca64b3f78e22191addc94","clusterId":"c_59f8b968edae83a55b4f36f6","slug":"google-releases-embeddinggemma-2-with-8k-context-and-740m-parameters","model":"deepseek-v4-1-flash","headline":"Google Releases EmbeddingGemma 2 With 8K Context And 740M Parameters","summary":"Google announced EmbeddingGemma 2, an embedding model that maps text, code, images, video, and audio into one vector space. It has 740 million parameters, an 8K-token context window, and output vectors that developers can shorten from 768 dimensions to 512, 256, or 128. Google says text-only weights need about 191MB of active RAM on a Pixel 11 Pro.","whyItMatters":"The pitch is local: a multimodal embedder small enough to hold its working memory to roughly 191MB for text on a phone is aimed at on-device search and retrieval-augmented generation, where the vector index sits on the hardware instead of a server.","webCardHtml":"\u003cp\u003eGoogle says the model runs locally on mobile and desktop hardware, not only in a data center. On a Pixel 11 Pro, the text-only weights need a minimum of about 191MB of active RAM, and the full multimodal version about 567MB. For text-only work the model runs at 270 million parameters; adding a 170 million parameter vision encoder and a 300 million parameter audio encoder gives full multimodal support.\u003c/p\u003e\u003cp\u003eIt is built on the Gemma 4 architecture and released under the Apache 2.0 license, which allows commercial use. Weights are available on Hugging Face and Kaggle, with the Gemini Enterprise Agent Platform Model Garden listed as a coming home.\u003c/p\u003e","blueskyPost":"Google's EmbeddingGemma 2 maps text, code, images, video and audio into one vector space. 740M parameters, 8K context, and text-only weights held to about 191MB of active RAM on a Pixel 11 Pro.","twitterPost":"Google announced EmbeddingGemma 2: 740M parameters, 8K token context, multimodal embeddings from text to video and audio. Text-only weights run at about 191MB of active RAM on a Pixel 11 Pro.","threadsPost":null,"newsletterBlurb":"Google announced EmbeddingGemma 2, an embedding model that puts text, code, images, video and audio into a single vector space. It carries 740 million parameters and an 8K token context window, and Google says text-only weights hold to about 191MB of active RAM on a Pixel 11 Pro. The weights are on Hugging Face and Kaggle under the Apache 2.0 license.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20261007-google-embeddinggemma-2/\",\"title\":\"Google's embedding model \\\"EmbeddingGemma 2\\\" arrives, runs on smartphones with as little as about 191MB of RAM\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4715,"outputTokens":634,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1791353242,"createdAt":"2026-10-07T06:03:04.000Z","publishedAt":"2026-10-07T06:06:10.000Z","updatedAt":"2026-10-07T06:06:10.000Z"},"cluster":{"id":"c_59f8b968edae83a55b4f36f6","canonicalTitle":"Googleの埋め込みモデル「EmbeddingGemma 2」が登場、使用RAMは最小約191MBでスマホでも動作","representativeArticleId":"a_7e3f876cb3a7719e1ece6889","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"EmbeddingGemma 2\"],\"studios\":[\"Google\"],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-10-07T04:47:00.000Z","lastSeenAt":"2026-10-07T04:47:00.000Z","updatedAt":"2026-10-07T06:06:06.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20261007-google-embeddinggemma-2/","title":"Googleの埋め込みモデル「EmbeddingGemma 2」が登場、使用RAMは最小約191MBでスマホでも動作"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["EmbeddingGemma 2"],"studios":["Google"],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":null}