NVIDIA Nemotron Achieves Best Benchmark Performance with LangChain Deep Agents Harness
NVIDIA Nemotron 3 Ultra offers better performance at a lower cost than high-end closed-loop models with a large and widely accepted AI agent platform.
LangChain tuned its Deep Agents harness for NVIDIA Nemotron 3 Ultra, achieving the highest accuracy among open models, while completing many tasks with higher output and operating at 10x lower cost per run than leading closed models.
Measured against LangChain's Deep Agents benchmark, Nemotron 3 Ultra also achieved business performance parity with the highest value closed models. No retraining of the model was required. All the benefit came from engineering the environment around the model, not the model itself.
For $10, teams using the NVIDIA Nemotron 3 Ultra can conduct continuous testing, quickly test and build specialized agents across their business.
LangChain's agent engineering platform has more than 200 million monthly downloads. By tuning its Deep Agents harness in particular NVIDIA Nemotron 3 Ultra, enables high-performance agents that multitask, run faster and provide businesses with a fully open stack that they can customize, own and work from anywhere.
“The way to build better agents is to continue to develop the system around the model,” said Harrison Chase, founder and CEO of LangChain. “Memory, tooling, testing and modeling behavior when teams can bring them together. Our work with NVIDIA shows that enterprises can achieve robust performance in an open stack while managing the agent systems they build.”
Abridge, Amdocs again The box they embed special agents in their platforms and global systems integrator EY expands its NVIDIA operating capabilities around the NVIDIA NemoClaw blueprints for LangChain Deep Agents, helping customers to customize, test and manage special agents across high-value workflows.
NVIDIA co-founder and CEO Jensen Huang recently sat down with Chase to discuss why the past six months have seen such a leap in AI for businesses.
Wiring Engineering, Not Fine Tuning
LangChain's team ran Nemotron 3 Ultra against its public benchmark suite Deep Agents, and analyzed deep agent It tracks to find out exactly where it dropped the points. Instead of retraining the model, the team tuned the harness around it – adjusting system information, tool definitions and middleware.
Every developer using LangChain Deep Agents with Nemotron 3 Ultra can use this today — the tuned profile is available directly through LangChain.
An Open Stack Built to Own You
NVIDIA NemoClaw for LangChain Deep Agents is an open source blueprint that packages this functionality for self-build businesses. special AI – model systemstools and working time – optimized for their workflow. Includes LangChain Deep Agents Code, unlocked for Nemotron 3 Ultra, and NVIDIA OpenShell secure runtime to perform agent actions safely.
An open model, open harness and open secure runtime means businesses own the full stack, end-to-end. They can customize it around the technology that differentiates their business, keep improving it and run it anywhere – their infrastructure, their cloud, their governance.
That distinction becomes more important as agents take on higher responsibilities. The shift from AI assistants answering questions to agents taking action within core systems is changing what businesses get from their AI.
NemoClaw for LangChain Deep Agents and Nemotron 3 Ultra tuned model profile available now. Developers can pull harnesses of Deep Agents tuned directly from LangChain, or use them LangChain's NemoClaw The Deep Agents blueprint as a starting point for building special agents from scratch.
How to Get Started
LangChain developers can access Nemotron 3 Ultra unlocked Baseten, Crusoe Cloud, DeepInfra, Fireworks, Nebius again Together the AI platforms, giving them a direct, predictable path to a tuned harness in production.
EY can help businesses start building their own special agents today, using this open source software stack.
Learn more about NVIDIA NemoClaw for LangChain Deep Agents and NVIDIA Nemotron.
Stay up to date with agent AI, NVIDIA Nemotron and more by subscribing to NVIDIA news, to join the communityand following NVIDIA AI enabled LinkedIn, Instagram, X again Facebook.
Check it out self-paced video tutorials and live streaming.
