NVIDIA open-sources Nemotron 3 Embed

// 7h agoMODEL RELEASE

NVIDIA open-sources Nemotron 3 Embed

NVIDIA has released Nemotron 3 Embed, a family of high-performance text embedding models designed to power semantic search and Retrieval-Augmented Generation (RAG) workflows. The release features a flagship 8B parameter model ranking highly on embedding benchmarks alongside an efficient 1B variant optimized for resource-constrained environments.

// ANALYSIS

NVIDIA is aggressively expanding its open-weights footprint to dominate the developer stack for Retrieval-Augmented Generation (RAG).

–The top-ranking 8B flagship model delivers high accuracy for complex semantic retrieval tasks.
–The 1B variant addresses the critical market need for low-latency, resource-efficient vector search deployments.
–Releasing these models on Hugging Face encourages developers to build on NVIDIA's software stack, further locking them into their hardware ecosystem.

// TAGS

nvidianemotron-3-embedembedding-modelsopen-sourceragsemantic-searchai-modelshugging-face

DISCOVERED

7h ago

2026-07-19

PUBLISHED

7h ago

2026-07-19

RELEVANCE

9/ 10

AUTHOR

AI Search

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

INFRA26m ago

Japan builds 140MW Vera Rubin AI factory

Japan's Ministry of Economy, Trade and Industry has partnered with NVIDIA and consortium Noetra Corp to build a 140-megawatt Vera Rubin AI factory under the FRONTia Project. The state-backed infrastructure will train open multimodal foundation models for physical AI and robotics, distributing weights directly to domestic developers.

UPDATE51m ago

Mercury Agent introduces centralized session management

Mercury Agent, an open-source and persistent AI agent framework, has announced upcoming support for centralized session management. The new feature will enable users to start a session on one platform and seamlessly switch or resume it on another without losing context or restarting the conversation, using commands like /session new and /session switch to manage states across platforms such as terminal, Discord, Slack, Telegram, and Signal.

MODEL2h ago

Qwen-3.8-Max Outperforms GPT-5.6 Sol, Rivals Fable 5

The shared social media announcement highlights that Alibaba's upcoming flagship model, Qwen-3.8-Max, reportedly outperforms OpenAI's GPT-5.6 Sol and trails Anthropic's Fable 5 by only a narrow margin. This benchmark performance positions Qwen-3.8-Max as a top-tier contender in the rapidly evolving frontier model landscape of 2026, challenging traditional leaders like OpenAI and Anthropic.

NVIDIA open-sources Nemotron 3 Embed