[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-embeddinggemma-2-multimodal-embeddings":3,"topics-all":38,"news-related-c53a7c3b-7a0c-4cc6-822e-aa0d66e01abe":57},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"c53a7c3b-7a0c-4cc6-822e-aa0d66e01abe","EmbeddingGemma 2 开源:740M 多模态嵌入跑进手机","谷歌开源 EmbeddingGemma 2:740M 参数、Apache 2.0 协议,五模态统一进一个嵌入空间,检索在手机本地完成。纯文本最小 270M 参数,MRL 截断最多省 6 倍存储,Pixel 11 Pro 全模态约占 567MB 内存;MTEB Code 提分 9.92,端侧 RAG 拼图补齐。","谷歌把多模态嵌入模型 EmbeddingGemma 2 直接开源放上了 Hugging Face:740M 参数、Apache 2.0 协议,目标是让检索这类「脏活」在手机本地完成,而不是把数据传上云端做向量化。\n\n## 一个模型,五个模态\n\nEmbeddingGemma 2 基于 Gemma 4 架构,把文本、代码、图像、音频、视频映射进同一个嵌入空间。官方博客给出的用例很具体:用文字从语音备忘录里找出某段视频剪辑,或在几小时录音里做文本检索,全部由这一个模型在本地完成。它的上一代 EmbeddingGemma 是纯文本模型,官方称累计下载已超过 2000 万次,主要用在端侧搜索和隐私优先的 RAG 管线里——这一代补上的正是多模态这块拼图。\n\n## 模块化裁剪:按需付内存\n\n这个模型最务实的设计是可拆:纯文本负载最小只要 270M 参数,视觉编码器 170M、音频编码器 300M 都是可选模块。借助 Matryoshka 表示学习(MRL),输出向量能从 768 维动态截断到 512、256 甚至 128 维,官方称本地向量库存储最多可省 6 倍。上下文窗口 8K token,是上一代的 4 倍,可容纳约 5.5 分钟音频、29 张图片或 58 帧视频。量化后在 Google Pixel 11 Pro 上,纯文本权重约占 191MB 活动内存,完整多模态约 567MB——单机可跑。\n\n## 跑分与生态\n\n官方数据显示,MTEB Code 得分从上一代的 68.76 提到 78.68,提升 9.92 分;官方称其在 sub-1B 多模态嵌入模型中同级领先,部分任务上超过两倍于其规模的专业模型——需要说明,这些均为厂商自报数字。生态比较完整:权重上了 Hugging Face 和 Kaggle,llama.cpp、Ollama、MLX、vLLM、SGLang 均有官方支持。它还与 Gemma 4 共享文本 tokenizer 和音频编码器,两模型同跑时总内存占用更低,「Gemma 4 生成 + EmbeddingGemma 2 检索」的端侧 RAG 组合意图非常明显。\n\n## 所以呢\n\n端侧 AI 的叙事里,生成模型一直是主角,但真正难啃的往往是检索:没有本地嵌入,所谓隐私优先的 RAG 就得把数据送上云端做向量化。谷歌这次把这块拼图补上了,而且用 Apache 2.0 而非受限协议,对端侧开发者是真金白银的让利。值得观察的是,其他端侧生态会不会跟进推出对标的开源嵌入模型——检索这一层,正在成为端侧 AI 竞争的新前沿。\n\n(来源:[Google 官方博客](https:\u002F\u002Fblog.google\u002Finnovation-and-ai\u002Ftechnology\u002Fdevelopers-tools\u002Fembeddinggemma-2\u002F))","https:\u002F\u002Fblog.google\u002Finnovation-and-ai\u002Ftechnology\u002Fdevelopers-tools\u002Fembeddinggemma-2\u002F","d884df39-706a-45db-b7ed-371f12e54f1f",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"8cf7490f-2449-4ba7-be19-61befa0d92b4","google",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"id":19,"name":20,"slug":20,"description":14,"color":14},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",{"id":22,"name":23,"slug":23,"description":14,"color":14},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"62dd5771-cc78-4499-bfb8-45cc43fb73d6","en","EmbeddingGemma 2: 740M Multimodal Embeddings, On-Device","Google open-sources EmbeddingGemma 2 (740M, Apache 2.0): one embedding space for text, code, image, audio and video, on-device. MTEB Code +9.92.","Google has open-sourced EmbeddingGemma 2, its multimodal embedding model, straight to Hugging Face: 740M parameters under Apache 2.0, built to keep retrieval workloads on phones instead of shipping data to the cloud for vectorization.\n\n## One Model, Five Modalities\n\nBuilt on the Gemma 4 architecture, EmbeddingGemma 2 maps text, code, images, audio, and video into a single shared embedding space. The official blog gives concrete use cases: find a specific video clip from a voice memo, or search hours of audio recordings with a text query — all processed locally by one natively multimodal model. Its predecessor, EmbeddingGemma, was text-only; Google says it has passed 20 million downloads, mostly powering on-device search and privacy-first RAG pipelines. Multimodal support is exactly the piece this generation adds.\n\n## Modular by Design: Pay Memory Only for What You Use\n\nThe most pragmatic design choice is modularity: text-only workloads need as little as 270M parameters, with optional vision (170M) and audio (300M) encoders for full multimodal support. Using Matryoshka Representation Learning (MRL), output vectors can be dynamically truncated from 768 dimensions down to 512, 256, or 128 — Google claims up to 6x storage reduction for local vector databases. The context window is 8K tokens, 4x larger than EmbeddingGemma 1, fitting about 5.5 minutes of audio, 29 images, or 58 video frames. With quantization on a Google Pixel 11 Pro, the model needs as little as ~191MB active RAM for text-only weights and ~567MB for the full multimodal stack — single-device territory.\n\n## Benchmarks and Ecosystem\n\nOfficial numbers show MTEB Code improving from 68.76 to 78.68, a 9.92-point jump over the previous generation. Google claims best-in-class scores among sub-1B multimodal embedders, outperforming some specialist models more than twice its size — note these are all vendor-reported figures. The ecosystem support is thorough: weights are on Hugging Face and Kaggle, with llama.cpp, Ollama, MLX, vLLM, and SGLang all supported from day one. It also shares its text tokenizer and audio encoder with Gemma 4, lowering the combined memory footprint when both run together — the intent of a \"Gemma 4 for generation + EmbeddingGemma 2 for retrieval\" on-device RAG stack is hard to miss.\n\n## So What\n\nIn the on-device AI narrative, generative models get the spotlight, but retrieval is the harder problem: without local embeddings, privacy-first RAG still has to send data to the cloud for vectorization. Google has now filled that gap — and with Apache 2.0 rather than a restricted license, a real concession to edge developers. The question worth watching is whether other edge ecosystems will follow with comparable open embedding models. The retrieval layer is becoming the new front line of on-device AI competition.\n\n(Source: [Google's official blog](https:\u002F\u002Fblog.google\u002Finnovation-and-ai\u002Ftechnology\u002Fdevelopers-tools\u002Fembeddinggemma-2\u002F))","embeddinggemma-2-multimodal-embeddings","2026-10-11T21:10:00Z","2026-10-11T21:09:44.039394Z","2026-10-11T21:09:44.039402Z",true,"agent",7,[39,48],{"slug":40,"tag_slug":40,"title_zh":41,"title_en":42,"intro_zh":43,"intro_en":44,"id":45,"is_active":35,"created_at":46,"modified_at":47},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":49,"tag_slug":49,"title_zh":50,"title_en":51,"intro_zh":52,"intro_en":53,"id":54,"is_active":35,"created_at":55,"modified_at":56},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":58},[59,64,69,74,79,84],{"id":60,"title":61,"news_slug":62,"published_at":63},"df4159eb-60c1-4267-835b-0932d2a1ac72","d1-3B 开源：零token决策，8毫秒应答","liquid-ai-d1-open-edge-decision-models","2026-10-11T19:14:50+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"36c71e4e-e398-4583-8710-1732aefff06a","OneStreamer:4B 流式模型先记再答,八榜最佳","onestreamer-4b-streaming-video-memory","2026-10-02T15:07:48+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"bbe8d55a-1069-42fa-a342-d944d53fdb4b","操作电脑的 27B 开源权重模型 Holo4:最强版禁商用","holo4-open-weight-computer-use","2026-10-02T13:12:01+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"dc61debd-1d77-4d5a-9d59-5b23c3da07de","蚂蚁开源Realtime-Venus：9B全双工模型边说边干活，三项续聊指标超GPT-4o","ant-realtime-venus-full-duplex-delegation","2026-09-30T23:10:53+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"d219ead5-ea5d-4aaa-835e-5ec3cf980856","Gemini Live Avatar:模型直出唇同步数字人","gemini-3-8-live-avatar","2026-09-26T17:06:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"0d357a0b-42da-40af-8038-e8c035cb9810","Apple 开源 LensVLM-9B:先扫压缩图,再读原页","apple-lensvlm-9b-weights-huggingface","2026-09-26T15:20:00+00:00"]