--- title: NVIDIA and LangChain release open agent blueprint with... description: NVIDIA and LangChain have released NemoClaw for LangChain Deep Agents, an open reference blueprint pairing LangChain's Deep Agents harness with NVIDIA's... canonical: https://daily.dev/posts/nvidia-and-langchain-release-open-agent-blueprint-with-nemotron-3-ultra-citing-10x-cost-advantage-o-za5zhukzo twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: NVIDIA and LangChain release open agent blueprint with Nemotron 3 Ultra, citing 10x cost advantage over closed models | daily.dev og:description: NVIDIA and LangChain have released NemoClaw for LangChain Deep Agents, an open reference blueprint pairing LangChain's Deep Agents harness with NVIDIA's... og:url: https://daily.dev/posts/nvidia-and-langchain-release-open-agent-blueprint-with-nemotron-3-ultra-citing-10x-cost-advantage-o-za5zhukzo og:image: https://api.daily.dev/og/posts/Za5ZhuKZo.png og:image:alt: NVIDIA and LangChain release open agent blueprint with Nemotron 3 Ultra, citing 10x cost advantage over closed models og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # NVIDIA and LangChain release open agent blueprint with Nemotron 3 Ultra, citing 10x cost advantage over closed models **[Collections](https://daily.dev/sources/collections)** · 3 min read · 1 upvotes · 0 comments ## Summary NVIDIA and LangChain have released NemoClaw for LangChain Deep Agents, an open reference blueprint pairing LangChain's Deep Agents harness with NVIDIA's Nemotron 3 Ultra model. The benchmark gains were achieved purely through harness engineering — tuning system prompts, tool descriptions, and middleware — with no model retraining, delivering benchmark-leading performance among open models at roughly 10x lower inference cost than top closed alternatives. Arcee AI is running inference at around 90 cents per million output tokens. The release is part of the broader Nemotron Coalition, which includes shared training data (over 10 trillion pre-training tokens), new datasets, and ecosystem partners like Arcee AI, Prime Intellect, and Unsloth. Real-world deployments include Abridge (clinical summarization), Glean (enterprise search), and Harvey (legal AI). LangChain also released a new version of OpenWiki for building personal knowledge bases from Gmail and browsing history. ## Content ## What's happening NVIDIA and LangChain have been quietly building something worth paying attention to. This week LangChain launched the NemoClaw Deep Agents blueprint — a tuned version of their Deep Agents harness paired with NVIDIA's Nemotron 3 Ultra (550B parameters) and NVIDIA OpenShell, an enterprise-ready secure runtime. The headline number: benchmark-leading performance among open models at roughly 10x lower inference cost than top closed alternatives. Arcee AI is reportedly running Nemotron at around 90 cents per million output tokens — about 20x cheaper than comparable closed frontier models. ## How the performance gains happened Here's the part I find genuinely interesting: none of the benchmark improvements came from retraining the model. The team got there entirely through harness engineering — adjusting system prompts, tool descriptions, and middleware. No fine-tuning, no new weights. Just careful prompt and tooling work on top of an already capable open model. That's a meaningful result. It suggests there's still a lot of headroom in how we deploy and configure these models, separate from the models themselves. ## The open data argument NVIDIA is also making a broader case that capable agents require more than open weights — they need open, inspectable training data. Their Nemotron open data ecosystem includes: - Over 10 trillion pre-training tokens - Millions of post-training samples - Nemotron Post-Training v3 Prompt Atlas (an interactive visual map of prompt data) - Nemotron-Personas: synthetically generated demographic personas now covering 2.4 billion people across 10 countries - Datasets including Nemotron-CC, Nemotron-MATH, and Nemotron-CLIMB The argument for synthetic data is practical: organizations can share useful training signals without exposing proprietary workflows. It also makes agent behavior easier to inspect and audit — something that matters a lot in regulated industries. ## Who's actually using this A few real deployments worth noting: - **Abridge** — clinical conversation summarization - **Glean** — agentic enterprise search - **Harvey** — legal AI, reportedly at 10x lower cost than their previous setup - **YTL AI Labs** — building a Malaysian language model on top of Nemotron On the tooling side, LangChain, Arcee AI, Prime Intellect, and Unsloth are all part of what NVIDIA is calling the Nemotron Coalition — a formalized ecosystem combining shared data, evaluations, and domain expertise. ## The sovereignty angle LangChain's Harrison Chase framed this week's launches around a theme that keeps coming up in enterprise AI conversations: owning your whole stack. Open models handle the model layer. Memory tools like OpenWiki (which just released a new version focused on "personal brains" built from Gmail, browsing history, etc.) handle the context layer. For companies that can't or won't send data to closed API providers — whether for legal, competitive, or regulatory reasons — a self-hostable stack from model weights to memory is genuinely useful. That's the pitch here, and it's a coherent one. ## Where to run it Nemotron 3 Ultra is available through Baseten, Fireworks, and Together AI for hosted inference. The NemoClaw blueprint is open and self-hostable for teams that want full control. ## Similar posts on daily.dev - [LangChain and NVIDIA Launch NemoClaw Deep Agents Blueprint](https://daily.dev/posts/langchain-and-nvidia-launch-nemoclaw-deep-agents-blueprint-ursnabiai) · LangChain · 2 upvotes · 0 comments --- Tags: [#open-source](https://daily.dev/tags/open-source), [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#nvidia](https://daily.dev/tags/nvidia), [#langchain](https://daily.dev/tags/langchain) [View this post on daily.dev](https://daily.dev/posts/nvidia-and-langchain-release-open-agent-blueprint-with-nemotron-3-ultra-citing-10x-cost-advantage-o-za5zhukzo) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"NVIDIA and LangChain release open agent blueprint with Nemotron 3 Ultra, citing 10x cost advantage over closed models","url":"https://daily.dev/posts/nvidia-and-langchain-release-open-agent-blueprint-with-nemotron-3-ultra-citing-10x-cost-advantage-o-za5zhukzo","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/nvidia-and-langchain-release-open-agent-blueprint-with-nemotron-3-ultra-citing-10x-cost-advantage-o-za5zhukzo"},"datePublished":"2026-07-14T18:13:15.944Z","dateModified":"2026-07-16T16:01:37.514Z","description":"NVIDIA and LangChain have released NemoClaw for LangChain Deep Agents, an open reference blueprint pairing LangChain's Deep Agents harness with NVIDIA's...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/beb82f58a630f835e2068a4d2cf1416d?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/beb82f58a630f835e2068a4d2cf1416d?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/nvidia-and-langchain-release-open-agent-blueprint-with-nemotron-3-ultra-citing-10x-cost-advantage-o-za5zhukzo","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,llm,ai-agents,nvidia,langchain","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"NVIDIA and LangChain release open agent blueprint with Nemotron 3 Ultra, citing 10x cost advantage over closed models"}]} ```