{"schema":"postcutoff/model@1","as_of":"2026-11-22T13:32:00+01:00","url":"https://postcutoff.com/m/embeddinggemma-2/","md":"https://postcutoff.com/m/embeddinggemma-2/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"embeddinggemma-2","name":"EmbeddingGemma 2","org":"Google DeepMind","family":"Gemma","released":"2026-10-06","status":"current","type":"embedding","modality_in":["text","image","video","audio"],"modality_out":["text"],"open_weights":true,"model_license":"apache-2.0","context_window":8192,"max_output":null,"knowledge_cutoff":null,"pricing":null,"price_line":"Price not published","access":[{"provider":"Hugging Face","model_id":"google/embeddinggemma-2","url":"https://huggingface.co/google/embeddinggemma-2"},{"provider":"Google docs","docs":"https://ai.google.dev/gemma/docs/embeddinggemma"}],"capabilities":[{"name":"Four modalities in one small open embedding space","detail":"Maps text (including code), images, video and audio into one 768-dimensional space with 740M parameters in total: a 270M text model (130M backbone + 140M embedder) plus optional 170M vision and 300M audio encoders that can be loaded separately.","first":false,"discovered":"launch","source":"https://huggingface.co/google/embeddinggemma-2"},{"name":"Matryoshka truncation","detail":"Embeddings can be cut to 512, 256 or 128 dimensions (up to 6x less storage); quality loss is small down to 256d.","first":false,"discovered":"launch","source":"https://huggingface.co/google/embeddinggemma-2"},{"name":"Big code-retrieval gain over v1","detail":"MTEB code v1 78.68 vs 68.76 for EmbeddingGemma 1; MTEB multilingual v2 61.36 vs 61.15; MMEB v2 overall 59.01 (Google's numbers).","first":false,"discovered":"launch","source":"https://huggingface.co/google/embeddinggemma-2"}],"entry":"2026-10-06-embeddinggemma-2","notes":"Announced 2026-10-06 on the Google blog. Apache 2.0 (Gemma 4 license), not gated on Hugging Face. Runs on phones and laptops; uses task-instruction prefixes (e.g. 'task: search result | query: ...') for text inputs. sentence-transformers id: google/embeddinggemma-2. No hosted Gemini API endpoint is listed; for hosted multimodal embeddings Google offers gemini-embedding-2.","verified":"2026-10-07","body_md":"Small open on-device embedding model for multimodal search, RAG, classification and clustering.\n\n```python\nfrom sentence_transformers import SentenceTransformer\nmodel = SentenceTransformer(\"google/embeddinggemma-2\")\nq = model.encode(\"What causes the northern lights?\", prompt_name=\"SearchQuery\")\nd = model.encode(\"The northern lights are caused by charged particles from the sun.\", prompt_name=\"Document\")\nprint(model.similarity(q, d))\n```\n\nSources: [Hugging Face model card](https://huggingface.co/google/embeddinggemma-2), [launch blog](https://blog.google/innovation-and-ai/technology/developers-tools/embeddinggemma-2/), [docs](https://ai.google.dev/gemma/docs/embeddinggemma).","page_url":"https://postcutoff.com/m/embeddinggemma-2/","standing":{"kind":"embedding","newer_count":0,"newer":[],"before":{"id":"gemini-embedding-2","name":"Gemini Embedding 2","org":"Google DeepMind","released":"2026-04-22","page_url":"https://postcutoff.com/m/gemini-embedding-2/"},"after":null,"weights":"public","licence":"apache-2.0","licence_permissive":true,"summary":"Weights are public under a permissive licence (Apache-2.0)."}}