{"apiVersion":"v1","taxonomyVersion":1,"generatedAt":"2026-10-29T14:06:34.818Z","data":{"recordId":"wsa_3t8yvtrgwdrt","score":0.996,"confidence":"high","capped":false,"capReason":null,"sources":[{"n":1,"url":"https://huggingface.co/google/embeddinggemma-2","publisher":"Google","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-29T00:08:40.326Z","detectedPublishedAt":"2026-09-14T14:53:42.000Z","rights":"original"},{"n":2,"url":"https://developers.googleblog.com/embeddinggemma-2-the-developer-guide/","publisher":"Google","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-29T00:08:40.326Z","detectedPublishedAt":"2026-10-06T00:00:00.000Z","rights":"original"},{"n":3,"url":"https://developers.googleblog.com/google-ai-edge-with-embeddinggemma-2/","publisher":"Google","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-29T00:08:40.326Z","detectedPublishedAt":"2026-10-06T00:00:00.000Z","rights":"original"},{"n":4,"url":"https://huggingface.co/EverPaper/embeddinggemma-2-coreai","publisher":"Hugging Face","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-29T00:08:40.326Z","detectedPublishedAt":"2026-10-12T08:34:13.000Z","rights":"original"}],"claims":[{"id":"a9a7fcc6-ffd9-40c3-b9a9-70b7ec7bc20b","kind":"capability","statement":"Supports more than 100 languages.","citations":[{"source":1,"supportProbability":0.96,"supports":true,"quote":null}]},{"id":"859c1a32-34a8-4dba-b9ac-4e5f09c7c596","kind":"deployment","statement":"It is designed to run on consumer hardware such as mobile devices and laptops.","citations":[{"source":1,"supportProbability":0.99,"supports":true,"quote":{"text":"Designed to run on consumer hardware such as mobile devices and laptops, EmbeddingGemma 2 delivers low-latency semantic representations for on-device applications, like search, retrieval-augmented generation (RAG), classification, and clustering.","rights":"original"}}]},{"id":"b7fdad9d-41f7-44fc-81e3-a895d6e4ee87","kind":"capability","statement":"EmbeddingGemma 2 maps text, code, images, video, and audio into a shared 768-dimensional space.","citations":[{"source":2,"supportProbability":0.98,"supports":true,"quote":{"text":"Based on Gemma 4, this sub-1B model maps text, code, images, video, and audio into a unified 768-dimensional space.","rights":"original"}}]},{"id":"55bdde37-ded4-4dd3-9ed5-dc6a422af286","kind":"capability","statement":"The model has configurations ranging from 270M parameters for text and code to 740M parameters for all modalities.","citations":[{"source":2,"supportProbability":0.98,"supports":true,"quote":{"text":"Its modular architecture lets you load only what you need, scaling from 270M parameters for text and code up to 740M parameters for all modalities.","rights":"original"}}]},{"id":"85978831-5320-4bc8-a70c-0c21349dcb07","kind":"capability","statement":"Matryoshka Representation Learning enables truncating embeddings from 768 dimensions to 128 dimensions.","citations":[{"source":2,"supportProbability":0.99,"supports":true,"quote":{"text":"Flexible vector storage: Matryoshka Representation Learning (MRL) enables dynamic truncation from 768 dimensions down to 128 dimensions, cutting vector database storage requirements while retaining much of the original quality.","rights":"original"}}]},{"id":"5d9a520d-9377-42c6-bc19-9f27fc9e5695","kind":"availability","statement":"Model weights are available on Hugging Face.","citations":[{"source":2,"supportProbability":0.96,"supports":true,"quote":{"text":"Download the Weights: Access the weights of the model (released under the Apache 2.0 license) directly on Hugging Face.","rights":"original"}},{"source":3,"supportProbability":0.74,"supports":true,"quote":null}]},{"id":"943244d0-f3b7-41d6-a9aa-e31f965bd158","kind":"deployment","statement":"The model can be run using sentence-transformers, Transformers, vLLM, SGLang, MLX, Ollama, LMStudio, and LiteRT.","citations":[{"source":2,"supportProbability":0.94,"supports":true,"quote":{"text":"Use Your Favorite Development Tools: Run the model efficiently using vLLM , Hugging Face Transformers , sentence-transformers , SGLang , MLX , Ollama , LMStudio , and LiteRT .","rights":"original"}}]},{"id":"f22f7d8c-1e08-4da3-a4b0-cd99fced65e0","kind":"recommendation-relevant","statement":"EmbeddingGemma 2 scores 14% higher than EmbeddingGemma 1 on MTEB (Code), according to the announcement.","citations":[{"source":2,"supportProbability":0.93,"supports":true,"quote":{"text":"EmbeddingGemma 2 scores 14% higher than EmbeddingGemma 1 on MTEB (Code), adds image, video, and audio retrieval while retaining the accuracy on multilingual text of EmbeddingGemma 1","rights":"original"}}]},{"id":"c4d897ce-f3da-4172-a452-3a2f60b4d1cb","kind":"capability","statement":"Text, image, and audio embeddings share a 768-dimensional space.","citations":[{"source":4,"supportProbability":0.96,"supports":true,"quote":{"text":"1.1 the image encoder and since 1.2 the audio encoder, in one shared 768-d space.","rights":"original"}}]},{"id":"14cdca16-b2f0-4a9e-8a27-dcdbd403b82f","kind":"capability","statement":"Audio embedding supports clips up to 30 seconds.","citations":[{"source":4,"supportProbability":0.98,"supports":true,"quote":{"text":"image-smoke-reference.json, preprocess-cases.json, check_preprocess.py, audio/ (16 generated clips","rights":"original"}}]},{"id":"e9309b7a-a047-4248-bfff-3f0adbddc1d6","kind":"capability","statement":"Audio encoding can run on the Neural Engine or GPU.","citations":[{"source":4,"supportProbability":0.99,"supports":true,"quote":{"text":"- **Two audio routes**: the audio encoder runs on the Neural Engine (default) or on the GPU, set independently of","rights":"original"}}]},{"id":"a5aba180-f800-4853-a556-a118a76d83b2","kind":"availability","statement":"The package can be downloaded from the EverPaper Hugging Face repository.","citations":[{"source":4,"supportProbability":0.91,"supports":true,"quote":null}]},{"id":"f065d0cf-4069-4660-9555-dadd09478a24","kind":"deployment","statement":"The package requires macOS 27 or later on Apple silicon.","citations":[{"source":4,"supportProbability":0.98,"supports":true,"quote":{"text":"embeddings) converted to Apple's **Core AI** runtime for **macOS 27 on Apple silicon**: the text encoder, since","rights":"original"}}]},{"id":"948a42fa-fc72-4275-8148-6f5ea44173b9","kind":"recommendation-relevant","statement":"The release is licensed under Apache-2.0 and points to the Gemma Prohibited Use Policy.","citations":[{"source":4,"supportProbability":0.97,"supports":true,"quote":null}]}]},"nextCursor":null,"license":{"catalog":{"name":"CC BY 4.0","spdx":"CC-BY-4.0","url":"https://creativecommons.org/licenses/by/4.0/","attribution":"wowsuchai","attributionUrl":"https://wowsuchai.withglint.app","covers":["catalog metadata","classifications","generated summaries","feed event metadata"]},"excluded":{"covers":["source quotations","linked source content"],"terms":"Not licensed by wowsuchai. Source quotations and linked source content keep their original rights."}}}