en.Wedoany.com Reported - LangChain has optimized its Deep Agents framework for the NVIDIA Nemotron 3 Ultra model, achieving the highest accuracy among open models while completing more tasks with higher throughput, with inference costs per run only one-tenth of leading closed-source models.

According to LangChain's Deep Agents benchmark, Nemotron 3 Ultra has reached the same level as high-scoring closed-source models on commercial tasks. All improvements come from engineering optimizations to the model's surrounding environment, not the model itself.

Teams using NVIDIA Nemotron 3 Ultra can run evaluations continuously, experiment faster, and build specialized agents across more business domains.

LangChain's agent engineering platform sees over 200 million downloads per month. By specifically tuning its Deep Agents framework for NVIDIA Nemotron 3 Ultra, the platform delivers high-performance agents that complete more tasks, run faster, and provide enterprises with a fully open stack.

"The way to build better agents is to continuously improve the system around the model," said Harrison Chase, co-founder and CEO of LangChain. "When teams can collaboratively tune memory, tool usage, evaluation, and model behavior, these elements compound. Our collaboration with NVIDIA shows that enterprises can achieve powerful performance from an open stack while maintaining control over the agent systems they are building."

Abridge, Amdocs, and Box are embedding specialized agents directly into their platforms, while global systems integrator EY is expanding its NVIDIA implementation capabilities around the NVIDIA NemoClaw blueprint for LangChain Deep Agents.

NVIDIA founder and CEO Jensen Huang recently met with Chase to discuss the progress made in AI available to enterprises over the past six months.

Framework engineering, not fine-tuning. The LangChain team ran Nemotron 3 Ultra against its publicly available Deep Agents benchmark suite, analyzing deep agent execution trajectories and pinpointing areas of loss. The team adjusted the framework around the model—system prompts, tool descriptions, and middleware—without retraining the model. The tuned configuration profile is directly available through LangChain.

The NVIDIA NemoClaw for LangChain Deep Agents is an open reference blueprint that encapsulates this work for enterprises building their own specialized AI. It combines LangChain Deep Agents code tuned for Nemotron 3 Ultra with the NVIDIA OpenShell security runtime for safely executing agent operations. An open model, open framework, and open security runtime mean enterprises can own the entire stack end-to-end and customize it around their own expertise.

NemoClaw for LangChain Deep Agents and the tuned Nemotron 3 Ultra model configuration profile are now available. LangChain developers can access Nemotron 3 Ultra on Baseten, Crusoe Cloud, DeepInfra, Fireworks, Nebius, and Together AI platforms, and EY can help enterprises build specialized agents using this open software stack.