{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T06:40:50.084Z","headline":"NVIDIA 发布 Nemotron 3 Embed 系列，8B 版本在 RTEB 基准上排名第一","description":"NVIDIA 发布 Nemotron 3 Embed 系列，包含三个开源 checkpoint，其中 8B-BF16 版本在 RTEB 基准上以 78.46 的平均 NDCG@10 排名第一。1B-NVFP4 版本在 Blackwell 上吞吐量比 BF16 高 2 倍，精度保留 99.5%，所有模型最大序列长度 32，768 tokens。","url":"https://www.aioga.com/news/cmronvv4105bdbitohe86a4un/","mainEntityOfPage":"https://www.aioga.com/news/cmronvv4105bdbitohe86a4un/","datePublished":"2026-07-17T07:53:02.000Z","dateModified":"2026-07-17T07:53:02.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.marktechpost.com/2026/07/17/nvidia-ai-releases-nemotron-3-embed-an-open-embedding-collection-whose-8b-checkpoint-ranks-1-on-rteb","https://aihot.virxact.com/items/cmronvv4105bdbitohe86a4un"],"canonicalUrl":"https://www.aioga.com/news/cmronvv4105bdbitohe86a4un/","directAnswer":{"@type":"Answer","text":"NVIDIA 发布 Nemotron 3 Embed 开源模型系列，共含三个 checkpoint。材料称，8B-BF16 在 RTEB 的16项公开任务上以平均 NDCG@10 78.46 排名第一；所有版本最大序列长度均为 32,768 tokens。","url":"https://www.aioga.com/news/cmronvv4105bdbitohe86a4un/","dateCreated":"2026-07-17T07:53:02.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"marktechpost.com source article","url":"https://www.marktechpost.com/2026/07/17/nvidia-ai-releases-nemotron-3-embed-an-open-embedding-collection-whose-8b-checkpoint-ranks-1-on-rteb","datePublished":"2026-07-17T07:53:02.000Z","provider":{"@type":"Organization","name":"marktechpost.com","url":"https://www.marktechpost.com/2026/07/17/nvidia-ai-releases-nemotron-3-embed-an-open-embedding-collection-whose-8b-checkpoint-ranks-1-on-rteb"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmronvv4105bdbitohe86a4un","datePublished":"2026-07-17T07:53:02.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmronvv4105bdbitohe86a4un"}}],"aggregationSource":"MarkTechPost（RSS）","originalPublisher":{"name":"marktechpost.com","url":"https://www.marktechpost.com/2026/07/17/nvidia-ai-releases-nemotron-3-embed-an-open-embedding-collection-whose-8b-checkpoint-ranks-1-on-rteb"},"article":{"id":"cmronvv4105bdbitohe86a4un","slug":"cmronvv4105bdbitohe86a4un","url":"https://www.aioga.com/news/cmronvv4105bdbitohe86a4un/","title":"NVIDIA 发布 Nemotron 3 Embed 系列，8B 版本在 RTEB 基准上排名第一","title_en":"NVIDIA AI Releases Nemotron 3 Embed： An Open Embedding Collection Whose 8B Checkpoint Ranks #1 on RTEB","summary":"NVIDIA 发布 Nemotron 3 Embed 系列，包含三个开源 checkpoint，其中 8B-BF16 版本在 RTEB 基准上以 78.46 的平均 NDCG@10 排名第一。1B-NVFP4 版本在 Blackwell 上吞吐量比 BF16 高 2 倍，精度保留 99.5%，所有模型最大序列长度 32，768 tokens。","source":"MarkTechPost（RSS）","sourceUrl":"https://www.marktechpost.com/2026/07/17/nvidia-ai-releases-nemotron-3-embed-an-open-embedding-collection-whose-8b-checkpoint-ranks-1-on-rteb","aiHotUrl":"https://aihot.virxact.com/items/cmronvv4105bdbitohe86a4un","publishedAt":"2026-07-17T07:53:02.000Z","category":"模型更新","score":70,"selected":true,"articleBody":["Embedding models decide which passages an agent ever sees. NVIDIA released Nemotron 3 Embed model：https://huggingface.co/collections/nvidia/nemotron-3-embed to work on that layer. It targets production-scale RAG, agentic retrieval, code retrieval, and agent memory.","The model collection includes three open checkpoints. Nemotron-3-Embed-8B-BF16 is the accuracy-first option. Nemotron-3-Embed-1B-BF16 carries the same design into a smaller footprint. Nemotron-3-Embed-1B-NVFP4 is the Blackwell-optimized 4-bit path.","All three are transformer encoders trained with bidirectional attention masking . The final embedding comes from average pooling over token-level representations. Maximum sequence length is 32,768 tokens on every checkpoint.","Each model was evaluated across 34 languages. All three carry the OpenMDW License Agreement, version 1.1 (OpenMDW-1.1) . Notably, the bases are Mistral models. The 8B is built with Ministral-3-8B-Instruct-2512 . Both 1B variants use Ministral-3-3B-Instruct-2512 .","Nemotron-3-Embed-8B-BF16 ranks #1 overall on RTEB：https://mteb-leaderboard.hf.space/benchmark/RTEB%28beta%29 (as of July 17 2026), the Retrieval Embedding Benchmark. Evaluation covers its 16 public tasks. Every figure below is average NDCG@10, at model sequence length 4096.","Two gaps are worth noting. The 1B gains 10.4 RTEB points over llama-nemotron-embed-vl-1b-v2 , the prior-generation baseline. Separately, NVFP4 costs 0.38 RTEB points against its BF16 parent, or 99.5% retention.","Those 1B scores come from a compression pipeline, not a smaller training run. The parent was nemotron-3-embed-3b , pruned and distilled across two iterative rounds.","First, the 3B parent was pruned to 2B using NVIDIA ModelOpt mcore_minitron Neural Architecture Search (NAS) . The search covers hidden width, FFN size, attention heads, and depth. It then picks the best candidate from the top-10 Pareto front. A 50k in-domain calibration corpus scored those candidates.","Next, the 2B model was distilled from the fine-tuned 8B embedding teacher. Distillation combined cosine distance loss (COS) and mean squared error (MSE) loss. The data blend was multilingual and in-domain. Finally, the same procedure repeated to produce the 1.14B checkpoint.","Compression then continues into the serving format. Quantization hit weights and activations of linear layers only, targeting the NVFP4 data type. The research team used nvidia-modelopt v0.45.0 . Quantization-Aware Distillation (QAD) followed, primarily to recover accuracy on long inputs.","Calibration used 512 samples: 256 queries and 256 passages from abisee/cnn_dailymail . QAD training used 20k samples.","The rsesearch team reports NVFP4 on Blackwell delivers up to 2x higher throughput than BF16 . It retains 99%+ of BF16 retrieval accuracy . The NVFP4 card also documents dynamic embedding sizes. You can slice the 2048-d vector from the start to 1024 or 512 dimensions. Re-normalize afterward.","Before touching code, watch the path run. It animates prefixing, bidirectional encoding, average pooling, L2 normalization, and dot-product scoring. Scores come from each card’s published expected output.","As that walkthrough implies, the checkpoints do not share runtime paths.","Alongside the checkpoints, NVIDIA research team released an optimized NIM microservice for the 1B model. The Rust-based NIM matches or outperforms the vLLM checkpoint on GB200 and RTX PRO 6000. NVIDIA tested input sequence lengths of 256 and 1024. Separately, NVIDIA NeMo AutoModel recipes cover fine-tuning and distillation.","With those paths in mind, prefixes come first. Queries take query: and documents take passage: . Embeddings are L2-normalized, so dot product equals cosine similarity.","encode_query and encode_document read the saved prompts. So you never add prefixes by hand. For serving, /v2/embed applies them from input_type instead:","Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us ：https://forms.gle/wbash1wF6efRj8G58","Asif Razzaq is the CEO of Marktechpost Media Inc.. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. His most recent endeavor is the launch of an Artificial Intelligence Media Platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is both technically sound and easily understandable by a wide audience. The platform boasts of over 2 million monthly views, illustrating its popularity among audiences."],"articleImages":[{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2019/06/Screen-Shot-2021-09-14-at-9.02.24-AM-300x300.png","alt":"","afterParagraph":17,"url":"/media/articles/cmronvv4105bdbitohe86a4un/787a6d54564e8e19.webp"},{"sourceUrl":"https://www.marktechpost.com/wp-content/uploads/2026/07/high-level-description-a-developer-focus_feCO3rqGV4ig6q7LaBcN2w_K5t5TwjZTPWy666HxX0epA-100x70.png","alt":"Patter SDK Guide to Building a Restaurant Booking Phone Agent with Dynamic Variables, Guardrails, Latency Dashboards, and Eval Checks","afterParagraph":18,"url":"/media/articles/cmronvv4105bdbitohe86a4un/c998049732b3b311.webp"}],"mediaStatus":"ok","articleBodyZh":["嵌入模型决定了代理能看到哪些段落。NVIDIA 发布了 Nemotron 3 Embed 模型：https://huggingface.co/collections/nvidia/nemotron-3-embed 用于在该层工作。它针对生产规模的 RAG、智能检索、代码检索和代理记忆。","模型集合包括三个开放检查点。Nemotron-3-Embed-8B-BF16 是优先考虑准确性的选项。Nemotron-3-Embed-1B-BF16 将相同设计压缩到更小的规模。Nemotron-3-Embed-1B-NVFP4 是经过 Blackwell 优化的 4 位路径。","这三者都是使用双向注意力掩码训练的 Transformer 编码器。最终嵌入来自对 token 级表示的平均池化。所有检查点的最大序列长度为 32,768 个 token。","每个模型在 34 种语言中进行了评估。三者均遵循 OpenMDW 许可证协议，第 1.1 版 (OpenMDW-1.1)。值得注意的是，基础模型为 Mistral 模型。8B 版本基于 Ministral-3-8B-Instruct-2512 构建。两个 1B 变体使用 Ministral-3-3B-Instruct-2512。","Nemotron-3-Embed-8B-BF16 在 RTEB 上总体排名第 1：https://mteb-leaderboard.hf.space/benchmark/RTEB(beta) （截止至 2026 年 7 月 17 日），即检索嵌入基准。评估覆盖其 16 个公开任务。下列每个数值均为 NDCG@10 的平均值，模型序列长度为 4096。","值得注意的两个差距。1B 在 RTEB 上比上一代基线 llama-nemotron-embed-vl-1b-v2 提升 10.4 分。另一方面，NVFP4 与其 BF16 父模型相比损失 0.38 RTEB 分，相当于保持率 99.5%。","这些 1B 分数来自压缩流程，而非更小规模的训练。父模型为 nemotron-3-embed-3b，经两轮迭代修剪和蒸馏得到。","首先，3B 父模型使用 NVIDIA ModelOpt mcore_minitron 神经架构搜索 (NAS) 修剪至 2B。搜索涵盖隐藏层宽度、FFN 大小、注意力头数和深度，然后从前 10 名 Pareto 前沿中选择最佳候选。50k 的域内校准语料对这些候选进行评分。","接下来，从微调后的 8B 嵌入教师模型蒸馏得到 2B 模型。蒸馏结合了余弦距离损失 (COS) 和均方误差损失 (MSE)。数据混合为多语言和域内数据。最后，同样的流程重复，生成 1.14B 检查点。","压缩然后继续进入服务格式。量化只影响线性层的权重和激活，目标数据类型为 NVFP4。研究团队使用了 nvidia-modelopt v0.45.0。随后进行了量化感知蒸馏（QAD），主要用于恢复长输入的准确性。","校准使用了 512 个样本：256 个查询和 256 个来自 abisee/cnn_dailymail 的段落。QAD 训练使用了 20k 个样本。","研究团队报告称，Blackwell 上的 NVFP4 提供的吞吐量比 BF16 高出最多 2 倍。它保留了 BF16 检索准确率的 99%。NVFP4 卡还支持动态嵌入大小。你可以将 2048 维向量从开头切到 1024 或 512 维度，之后进行重新归一化。","在操作代码之前，先观察路径运行。它会展示前缀添加、双向编码、平均池化、L2 归一化和点积评分。评分来自每个卡发布的预期输出。","正如该演示所示，检查点不共享运行时路径。","除了检查点外，NVIDIA 研究团队发布了针对 1B 模型的优化 NIM 微服务。基于 Rust 的 NIM 在 GB200 和 RTX PRO 6000 上与 vLLM 检查点匹配或表现更佳。NVIDIA 测试了输入序列长度为 256 和 1024。另有 NVIDIA NeMo AutoModel 配方涵盖微调和蒸馏。","考虑到这些路径，前缀先行。查询使用 query:，文档使用 passage:. 嵌入向量经过 L2 归一化，因此点积等于余弦相似度。","encode_query 和 encode_document 会读取保存的提示。因此你无需手动添加前缀。对于服务端，/v2/embed 会根据 input_type 应用前缀：","需要与我们合作推广你的 GitHub 仓库、Hugging Face 页面、产品发布或网络研讨会等？请与我们联系：https://forms.gle/wbash1wF6efRj8G58","Asif Razzaq 是 Marktechpost Media Inc. 的首席执行官。作为一位有远见的企业家和工程师，Asif 致力于利用人工智能的潜力造福社会。他最近的努力是推出人工智能媒体平台 Marktechpost，该平台因其对机器学习和深度学习新闻的深入报道而脱颖而出，既技术上可靠，又易于广大受众理解。该平台每月浏览量超过 200 万次，显示了其在观众中的受欢迎程度。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"NVIDIA 发布 Nemotron 3 Embed 开源模型系列，共含三个 checkpoint。材料称，8B-BF16 在 RTEB 的16项公开任务上以平均 NDCG@10 78.46 排名第一；所有版本最大序列长度均为 32,768 tokens。","background":"该系列面向生产级 RAG、智能体检索、代码检索与智能体记忆，采用双向注意力掩码的 Transformer 编码器，并对 token 表示进行平均池化。三个模型覆盖 34 种语言，使用 OpenMDW-1.1 许可。","viewpoint":"Aioga 判断，此次更新的重点并非只有榜单成绩，还包括 8B 精度优先、1B 较小体量与 NVFP4 面向 Blackwell 优化的分层选择。值得关注的是，公开成绩基于长度 4096 的评测，不能直接等同于 32,768 tokens 下的实际效果。","implications":"Aioga 判断，该系列可能为检索系统提供精度、模型体量与部署吞吐量之间的不同选择。材料称，1B-NVFP4 在 Blackwell 上吞吐量为 BF16 的 2 倍，并保留 99.5% 精度，但这一结果不应外推到其他硬件或业务负载。","nextStep":"值得关注的是后续独立复测，包括不同语言、代码与长文档检索场景，以及 32,768 tokens 输入下的质量、延迟和资源占用。同时应核对 OpenMDW-1.1 的具体使用条件，并验证 NVFP4 在目标 Blackwell 环境中的实际收益。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-07-23T00:48:32.342Z","sourceHash":"eabe68d8f86ed435","review":{"approved":true,"groundedness":95,"clarity":92,"duplicationRisk":18,"blockingIssues":[],"notes":["“三个模型覆盖 34 种语言”可更精确地表述为“三个模型均接受了涵盖 34 种语言的评测”，因为来源明确陈述的是评测范围。","“保留 99.5% 精度”具体指相对 BF16 父模型的 RTEB 分数保留率；如需避免泛化，可改为“保留 99.5% 的 RTEB 得分”。","将该系列称为“开源模型系列”与来源材料一致，但后续核对 OpenMDW-1.1 的具体授权条件是合理且必要的提醒。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","low-source-overlap","no-html","independent-ai-review"]}},"tags":["模型更新","MarkTechPost（RSS）"],"translations":{"zh-CN":{"title":"NVIDIA 发布 Nemotron 3 Embed 系列，8B 版本在 RTEB 基准上排名第一","summary":"NVIDIA 发布 Nemotron 3 Embed 系列，包含三个开源 checkpoint，其中 8B-BF16 版本在 RTEB 基准上以 78.46 的平均 NDCG@10 排名第一。1B-NVFP4 版本在 Blackwell 上吞吐量比 BF16 高 2 倍，精度保留 99.5%，所有模型最大序列长度 32，768 tokens。","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA 发布 Nemotron 3 Embed 系列，8B 版本在 RTEB 基准上排名第一 - Aioga AI资讯","description":"NVIDIA 发布 Nemotron 3 Embed 系列，包含三个开源 checkpoint，其中 8B-BF16 版本在 RTEB 基准上以 78.46 的平均 NDCG@10 排名第一。1B-NVFP4 版本在 Blackwell 上吞吐量比 BF16 高 2 倍，精度保留 99.5%，所有模型最大序列长度 32，768 tokens。","url":"https://www.aioga.com/news/cmronvv4105bdbitohe86a4un/"},"en":{"title":"NVIDIA released the Nemotron 3 Embed series, with the 8B version ranking first in the RTEB benchmark","summary":"NVIDIA released the Nemotron 3 Embed series, which includes three open-source checkpoints, with the 8B-BF16 version ranking first on the RTEB benchmark with an average NDCG@10 of 78.46. The 1B-NVFP4 version delivers twice the throughput of the BF16 on Blackwell, retains 99.5% accuracy, and has a maximum sequence length of 32,768 tokens across all models.","category":"Models","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA released the Nemotron 3 Embed series, with the 8B version ranking first in the RTEB benchmark - Aioga AI News","description":"NVIDIA released the Nemotron 3 Embed series, which includes three open-source checkpoints, with the 8B-BF16 version ranking first on the RTEB benchmark with an average NDCG@10 of 7","url":"https://www.aioga.com/en/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-18T04:59:03.284Z"},"ja":{"title":"NVIDIAはNemotron 3 Embedシリーズをリリースし、8BバージョンがRTEBベンチマークで1位にランクされました","summary":"NVIDIAはNemotron 3 Embedシリーズをリリースしました。これには3つのオープンソースチェックポイントが含まれており、8B-BF16版はRTEBベンチマークで平均78.46の平均NDCG@10で1位にランクされました。 1B-NVFP4バージョンはBlackwellのBF16の2倍のスループットを提供し、99.5%の精度を維持し、全モデルで最大32,768トークンのシーケンス長を持ちます。","category":"モデル更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIAはNemotron 3 Embedシリーズをリリースし、8BバージョンがRTEBベンチマークで1位にランクされました - Aioga AIニュース","description":"NVIDIAはNemotron 3 Embedシリーズをリリースしました。これには3つのオープンソースチェックポイントが含まれており、8B-BF16版はRTEBベンチマークで平均78.46の平均NDCG@10で1位にランクされました。 1B-NVFP4バージョンはBlackwellのBF16の2倍のスループットを提供し、99.5%の精度を維持し、全モデルで最","url":"https://www.aioga.com/ja/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-18T04:59:22.116Z"},"ko":{"title":"NVIDIA는 Nemotron 3 Embed 시리즈를 출시했으며, 8B 버전이 RTEB 벤치마크에서 1위를 차지했습니다","summary":"NVIDIA는 세 개의 오픈 소스 체크포인트를 포함하는 Nemotron 3 Embed 시리즈를 출시했으며, 8B-BF16 버전은 RTEB 벤치마크에서 평균 NDCG@10 78.46으로 1위를 차지했습니다. 1B-NVFP4 버전은 Blackwell의 BF16보다 두 배의 처리량을 제공하며, 정확도는 99.5%를 유지하고, 모든 모델에서 최대 32,768개의 토큰 시퀀스 길이를 가집니다.","category":"모델 업데이트","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA는 Nemotron 3 Embed 시리즈를 출시했으며, 8B 버전이 RTEB 벤치마크에서 1위를 차지했습니다 - Aioga AI 뉴스","description":"NVIDIA는 세 개의 오픈 소스 체크포인트를 포함하는 Nemotron 3 Embed 시리즈를 출시했으며, 8B-BF16 버전은 RTEB 벤치마크에서 평균 NDCG@10 78.46으로 1위를 차지했습니다. 1B-NVFP4 버전은 Blackwell의 BF16보다 두 배의 처리량을 제공하며, 정확도는 99.5%를 유지하고, ","url":"https://www.aioga.com/ko/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-18T04:59:41.853Z"},"es":{"title":"NVIDIA lanzó la serie Nemotron 3 Embed, con la versión 8B ocupando el primer puesto en el benchmark RTEB","summary":"NVIDIA lanzó la serie Nemotron 3 Embed, que incluye tres puntos de control de código abierto, siendo la versión 8B-BF16 la primera en el benchmark RTEB con un NDCG@10 medio de 78,46. La versión 1B-NVFP4 ofrece el doble de rendimiento que el BF16 en Blackwell, mantiene un 99,5% de precisión y una longitud máxima de secuencia de 32.768 tokens en todos los modelos.","category":"Modelos","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA lanzó la serie Nemotron 3 Embed, con la versión 8B ocupando el primer puesto en el benchmark RTEB - Aioga Noticias de IA","description":"NVIDIA lanzó la serie Nemotron 3 Embed, que incluye tres puntos de control de código abierto, siendo la versión 8B-BF16 la primera en el benchmark RTEB con un NDCG@10 medio de 78,4","url":"https://www.aioga.com/es/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-18T05:00:02.130Z"},"fr":{"title":"NVIDIA a sorti la série Nemotron 3 Embed, la version 8B se classant première dans le benchmark RTEB","summary":"NVIDIA a publié la série Nemotron 3 Embed, qui comprend trois points de contrôle open source, la version 8B-BF16 se classant première au benchmark RTEB avec un NDCG@10 moyen de 78,46. La version 1B-NVFP4 offre un débit deux fois supérieur au BF16 sur Blackwell, conserve une précision de 99,5 % et a une longueur de séquence maximale de 32 768 jetons pour tous les modèles.","category":"Modèles","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA a sorti la série Nemotron 3 Embed, la version 8B se classant première dans le benchmark RTEB - Aioga Actualités IA","description":"NVIDIA a publié la série Nemotron 3 Embed, qui comprend trois points de contrôle open source, la version 8B-BF16 se classant première au benchmark RTEB avec un NDCG@10 moyen de 78,","url":"https://www.aioga.com/fr/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-18T05:00:21.395Z"},"de":{"title":"NVIDIA brachte die Nemotron 3 Embed-Serie heraus, wobei die 8B-Version im RTEB-Benchmark den ersten Platz belegte","summary":"NVIDIA veröffentlichte die Nemotron 3 Embed-Serie, die drei Open-Source-Checkpoints enthält, wobei die 8B-BF16-Version mit einem durchschnittlichen NDCG@10 von 78,46 den ersten Platz im RTEB-Benchmark belegte. Die 1B-NVFP4-Version liefert auf Blackwell den doppelten Durchsatz des BF16, behält 99,5 % Genauigkeit und hat eine maximale Sequenzlänge von 32.768 Token über alle Modelle hinweg.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA brachte die Nemotron 3 Embed-Serie heraus, wobei die 8B-Version im RTEB-Benchmark den ersten Platz belegte - Aioga KI-News","description":"NVIDIA veröffentlichte die Nemotron 3 Embed-Serie, die drei Open-Source-Checkpoints enthält, wobei die 8B-BF16-Version mit einem durchschnittlichen NDCG@10 von 78,46 den ersten Pla","url":"https://www.aioga.com/de/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:55.358Z"},"pt-BR":{"title":"A NVIDIA lançou a série Nemotron 3 Embed, com a versão 8B ocupando o primeiro lugar no benchmark RTEB","summary":"A NVIDIA lançou a série Nemotron 3 Embed, que inclui três checkpoints open-source, com a versão 8B-BF16 ocupando o primeiro lugar no benchmark RTEB com uma média de NDCG@10 de 78,46. A versão 1B-NVFP4 entrega o dobro do débito do BF16 na Blackwell, mantém 99,5% de precisão e possui um comprimento máximo de sequência de 32.768 tokens em todos os modelos.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"A NVIDIA lançou a série Nemotron 3 Embed, com a versão 8B ocupando o primeiro lugar no benchmark RTEB - Aioga Notícias de IA","description":"A NVIDIA lançou a série Nemotron 3 Embed, que inclui três checkpoints open-source, com a versão 8B-BF16 ocupando o primeiro lugar no benchmark RTEB com uma média de NDCG@10 de 78,4","url":"https://www.aioga.com/pt-BR/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:55.819Z"},"ru":{"title":"NVIDIA выпустила серию Nemotron 3 Embed, при этом версия 8B заняла первое место в бенчмарке RTEB","summary":"NVIDIA выпустила серию Nemotron 3 Embed, включающую три контрольных пункта с открытым исходным кодом, при этом версия 8B-BF16 заняла первое место в бенчмарке RTEB со средним NDCG@10 78,46. Версия 1B-NVFP4 обеспечивает вдвое большую пропускную способность BF16 на Blackwell, сохраняет точность 99,5% и имеет максимальную длину последовательности 32 768 токенов на всех моделях.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA выпустила серию Nemotron 3 Embed, при этом версия 8B заняла первое место в бенчмарке RTEB - Aioga Новости ИИ","description":"NVIDIA выпустила серию Nemotron 3 Embed, включающую три контрольных пункта с открытым исходным кодом, при этом версия 8B-BF16 заняла первое место в бенчмарке RTEB со средним NDCG@1","url":"https://www.aioga.com/ru/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:56.701Z"},"ar":{"title":"أصدرت NVIDIA سلسلة نيموترون 3 إميبد، حيث احتلت نسخة 8B المرتبة الأولى في اختبار RTEB","summary":"أصدرت NVIDIA سلسلة Nemotron 3 Embed، التي تتضمن ثلاث نقاط تفتيش مفتوحة المصدر، حيث احتل إصدار 8B-BF16 المرتبة الأولى في اختبار RTEB بمتوسط NDCG@10 بلغ 78.46. تقدم نسخة 1B-NVFP4 ضعف معدل النقل مقارنة ب BF16 على بلاكويل، وتحافظ على دقة 99.5٪، ويبلغ طول التسلسل الأقصى 32,768 رمزا عبر جميع الطرازات.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"أصدرت NVIDIA سلسلة نيموترون 3 إميبد، حيث احتلت نسخة 8B المرتبة الأولى في اختبار RTEB - Aioga أخبار الذكاء الاصطناعي","description":"أصدرت NVIDIA سلسلة Nemotron 3 Embed، التي تتضمن ثلاث نقاط تفتيش مفتوحة المصدر، حيث احتل إصدار 8B-BF16 المرتبة الأولى في اختبار RTEB بمتوسط NDCG@10 بلغ 78.46. تقدم نسخة 1B-NVFP4 ضعف","url":"https://www.aioga.com/ar/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:56.688Z"},"hi":{"title":"NVIDIA ने नेमोट्रॉन 3 एम्बेड श्रृंखला जारी की, जिसमें 8B संस्करण RTEB बेंचमार्क में पहले स्थान पर है","summary":"NVIDIA ने नेमोट्रॉन 3 एम्बेड श्रृंखला जारी की, जिसमें तीन ओपन-सोर्स चेकपॉइंट शामिल हैं, जिसमें 8B-BF16 संस्करण 78.46 की औसत NDCG@10 के साथ RTEB बेंचमार्क पर पहले स्थान पर है। 1B-NVFP4 संस्करण ब्लैकवेल पर BF16 के थ्रूपुट से दोगुना थ्रूपुट प्रदान करता है, 99.5% सटीकता बरकरार रखता है, और सभी मॉडलों में इसकी अधिकतम अनुक्रम लंबाई 32,768 टोकन है।","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA ने नेमोट्रॉन 3 एम्बेड श्रृंखला जारी की, जिसमें 8B संस्करण RTEB बेंचमार्क में पहले स्थान पर है - Aioga AI समाचार","description":"NVIDIA ने नेमोट्रॉन 3 एम्बेड श्रृंखला जारी की, जिसमें तीन ओपन-सोर्स चेकपॉइंट शामिल हैं, जिसमें 8B-BF16 संस्करण 78.46 की औसत NDCG@10 के साथ RTEB बेंचमार्क पर पहले स्थान पर है। 1B-NV","url":"https://www.aioga.com/hi/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:56.433Z"},"it":{"title":"NVIDIA ha rilasciato la serie Nemotron 3 Embed, con la versione 8B che si è classificata prima nel benchmark RTEB","summary":"NVIDIA ha rilasciato la serie Nemotron 3 Embed, che include tre checkpoint open-source, con la versione 8B-BF16 che si è classificata prima nel benchmark RTEB con una NDCG@10 media di 78,46. La versione 1B-NVFP4 offre il doppio della produttività del BF16 su Blackwell, mantiene una precisione del 99,5% e una lunghezza massima della sequenza di 32.768 token su tutti i modelli.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA ha rilasciato la serie Nemotron 3 Embed, con la versione 8B che si è classificata prima nel benchmark RTEB - Aioga Notizie IA","description":"NVIDIA ha rilasciato la serie Nemotron 3 Embed, che include tre checkpoint open-source, con la versione 8B-BF16 che si è classificata prima nel benchmark RTEB con una NDCG@10 media","url":"https://www.aioga.com/it/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:55.786Z"},"nl":{"title":"NVIDIA bracht de Nemotron 3 Embed-serie uit, waarbij de 8B-versie als eerste eindigde in de RTEB-benchmark","summary":"NVIDIA bracht de Nemotron 3 Embed-serie uit, die drie open-source checkpoints bevat, waarbij de 8B-BF16 versie eerste staat op de RTEB-benchmark met een gemiddelde NDCG@10 van 78,46. De 1B-NVFP4 versie levert twee keer zoveel doorvoer als de BF16 op Blackwell, behoudt 99,5% nauwkeurigheid en heeft een maximale sequentielengte van 32.768 tokens over alle modellen.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA bracht de Nemotron 3 Embed-serie uit, waarbij de 8B-versie als eerste eindigde in de RTEB-benchmark - Aioga AI-nieuws","description":"NVIDIA bracht de Nemotron 3 Embed-serie uit, die drie open-source checkpoints bevat, waarbij de 8B-BF16 versie eerste staat op de RTEB-benchmark met een gemiddelde NDCG@10 van 78,4","url":"https://www.aioga.com/nl/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:55.790Z"},"tr":{"title":"NVIDIA, Nemotron 3 Embed serisini piyasaya sürdü; 8B versiyonu RTEB benchmarkında birinci sırada yer aldı","summary":"NVIDIA, üç açık kaynak kontrol noktası içeren Nemotron 3 Embed serisini piyasaya sürdü; 8B-BF16 versiyonu RTEB benchmarkında ortalama 78,46 NDCG@10 ile birinci sırada yer aldı. 1B-NVFP4 versiyonu, Blackwell'te BF16'nın iki katı verimliliği sağlar, %99,5 doğruluk sağlar ve tüm modellerde maksimum dizi uzunluğu 32.768 jeton olarak belirlenir.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA, Nemotron 3 Embed serisini piyasaya sürdü; 8B versiyonu RTEB benchmarkında birinci sırada yer aldı - Aioga AI Haberleri","description":"NVIDIA, üç açık kaynak kontrol noktası içeren Nemotron 3 Embed serisini piyasaya sürdü; 8B-BF16 versiyonu RTEB benchmarkında ortalama 78,46 NDCG@10 ile birinci sırada yer aldı. 1B-","url":"https://www.aioga.com/tr/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:55.987Z"},"vi":{"title":"NVIDIA đã phát hành dòng Nemotron 3 Embed, với phiên bản 8B đứng đầu trong điểm chuẩn RTEB","summary":"NVIDIA đã phát hành dòng Nemotron 3 Embed, bao gồm ba điểm kiểm tra mã nguồn mở, với phiên bản 8B-BF16 đứng đầu trên điểm chuẩn RTEB với NDCG@10 trung bình là 78,46. Phiên bản 1B-NVFP4 cung cấp thông lượng gấp đôi BF16 trên Blackwell, giữ được độ chính xác 99,5% và có độ dài trình tự tối đa là 32.768 mã thông báo trên tất cả các kiểu máy.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA đã phát hành dòng Nemotron 3 Embed, với phiên bản 8B đứng đầu trong điểm chuẩn RTEB - Tin tức AI Aioga","description":"NVIDIA đã phát hành dòng Nemotron 3 Embed, bao gồm ba điểm kiểm tra mã nguồn mở, với phiên bản 8B-BF16 đứng đầu trên điểm chuẩn RTEB với NDCG@10 trung bình là 78,46. Phiên bản 1B-N","url":"https://www.aioga.com/vi/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:56.408Z"},"id":{"title":"NVIDIA merilis seri Nemotron 3 Embed, dengan versi 8B menempati peringkat pertama dalam benchmark RTEB","summary":"NVIDIA merilis seri Nemotron 3 Embed, yang mencakup tiga pos pemeriksaan sumber terbuka, dengan versi 8B-BF16 menempati peringkat pertama di benchmark RTEB dengan NDCG@10 rata-rata 78,46. Versi 1B-NVFP4 memberikan throughput dua kali lipat dari BF16 di Blackwell, mempertahankan akurasi 99,5%, dan memiliki panjang urutan maksimum 32.768 token di semua model.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA merilis seri Nemotron 3 Embed, dengan versi 8B menempati peringkat pertama dalam benchmark RTEB - Berita AI Aioga","description":"NVIDIA merilis seri Nemotron 3 Embed, yang mencakup tiga pos pemeriksaan sumber terbuka, dengan versi 8B-BF16 menempati peringkat pertama di benchmark RTEB dengan NDCG@10 rata-rata","url":"https://www.aioga.com/id/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:55.444Z"},"th":{"title":"NVIDIA เปิดตัวซีรีส์ Nemotron 3 Embed โดยเวอร์ชัน 8B อยู่ในอันดับแรกในเกณฑ์มาตรฐาน RTEB","summary":"NVIDIA เปิดตัวซีรีส์ Nemotron 3 Embed ซึ่งรวมถึงจุดตรวจโอเพ่นซอร์สสามจุด โดยเวอร์ชัน 8B-BF16 อยู่ในอันดับแรกในเกณฑ์มาตรฐาน RTEB ด้วยค่า NDCG@10 เฉลี่ย 78.46 เวอร์ชัน 1B-NVFP4 ให้ปริมาณงานเป็นสองเท่าของ BF16 บน Blackwell รักษาความแม่นยํา 99.5% และมีความยาวลําดับสูงสุด 32,768 โทเค็นในทุกรุ่น","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA เปิดตัวซีรีส์ Nemotron 3 Embed โดยเวอร์ชัน 8B อยู่ในอันดับแรกในเกณฑ์มาตรฐาน RTEB - ข่าว AI Aioga","description":"NVIDIA เปิดตัวซีรีส์ Nemotron 3 Embed ซึ่งรวมถึงจุดตรวจโอเพ่นซอร์สสามจุด โดยเวอร์ชัน 8B-BF16 อยู่ในอันดับแรกในเกณฑ์มาตรฐาน RTEB ด้วยค่า NDCG@10 เฉลี่ย 78.46 เวอร์ชัน 1B-NVFP4 ให้ปร","url":"https://www.aioga.com/th/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:55.962Z"},"pl":{"title":"NVIDIA wypuściła serię Nemotron 3 Embed, a wersja 8B zajęła pierwsze miejsce w benchmarku RTEB","summary":"NVIDIA wypuściła serię Embed Nemotron 3, która zawiera trzy otwarte punkty kontrolne, a wersja 8B-BF16 zajęła pierwsze miejsce w benchmarku RTEB ze średnią NDCG@10 78,46. Wersja 1B-NVFP4 zapewnia dwukrotnie większą przepustowość niż BF16 w Blackwell, zachowuje dokładność 99,5% i ma maksymalną długość sekwencji 32 768 tokenów we wszystkich modelach.","category":"模型更新","source":"MarkTechPost（RSS）","aggregationSource":"MarkTechPost（RSS）","pageTitle":"NVIDIA wypuściła serię Nemotron 3 Embed, a wersja 8B zajęła pierwsze miejsce w benchmarku RTEB - Aioga Wiadomości AI","description":"NVIDIA wypuściła serię Embed Nemotron 3, która zawiera trzy otwarte punkty kontrolne, a wersja 8B-BF16 zajęła pierwsze miejsce w benchmarku RTEB ze średnią NDCG@10 78,46. Wersja 1B","url":"https://www.aioga.com/pl/news/cmronvv4105bdbitohe86a4un/","contentTranslated":true,"sourceHash":"48a58076c16f924e","translatedAt":"2026-07-19T15:57:57.047Z"}}}}