Decision Optimization
NVIDIA cuOpt
Achieve world-record speed on large-scale problems with millions of constraints and variables—saving time and reducing costs. Extend to agentic workflows with cuOpt agent skills.
Overview
What Is NVIDIA cuOpt?
NVIDIA® cuOpt™ is an open source, GPU-accelerated engine for decision optimization, excelling in mixed-integer programming (MIP), linear programming (LP), vehicle routing problems (VRPs), and quadratic programming (QP). Designed to tackle large-scale problems with millions of variables and constraints, cuOpt enables accelerated decision-making.
cuOpt seamlessly integrates into agentic workflows through open source cuOpt agent skills that help AI agents formulate, solve, debug, and explain optimization problems.
Benefits
Benefits of NVIDIA cuOpt GPU-Accelerated Optimization
GPU-Powered Speedup for Optimization Solvers
Enjoy significant speedups when lower-accuracy solutions are acceptable. Outperform commercial state-of-the-art VRP solvers.
Achieve World-Record Vehicle Routing Solutions
Achieve a world-record solution validated on the MIPLIB open problem, competitive performance on large LPs demonstrated by the Mittelmann benchmarks, and unmatched precision for VRP, validated by the Gehring & Homberger and Li & Lim benchmarks.
Dynamic and Batch Optimization
Continuously adapt to changing variables and constraints by rerunning models in near real time or batch mode for optimal decision-making.
Flexible Integration for Stand-Alone and Agentic Workflows
Use out of the box or seamlessly embed into your solver and/or agentic workflows for unmatched speed, scalability, and accuracy.
Enterprise Support With NVIDIA AI Enterprise
Accelerate time to value with the security, reliability, and enterprise-class support of NVIDIA AI Enterprise for production deployments.
Use Cases
How cuOpt Is Being Used
Explore how NVIDIA cuOpt powers real-world industry use cases, and jump-start your AI development with curated examples.
-
Supply Chain Management
-
Fleet Management
-
Last-Mile Delivery
-
Field Dispatch
-
Job Scheduling Optimization
-
Portfolio Optimization
Supply Chain Management
Optimizing resource allocation in complex supply chains requires efficiently distributing limited resources while adapting to real-time changes. With countless variables at play, achieving maximum productivity and cost efficiency demands rapid, intelligent decision-making. NVIDIA’s cuOpt-powered AI agent enables you to talk to your supply chain data via NVIDIA NIM™, delivering real-time, optimal resource allocation for greater operational agility and optimizing your resource allocation.
Fleet Management
Efficient scheduling and route planning are essential for managing inbound and outbound transportation of goods and vehicles, especially for long-haul fleets.
NVIDIA cuOpt, integrated with Omniverse™ Digital Twins, optimizes logistics by simulating real-world fleet operations in a virtual environment, enabling dynamic scheduling, route optimization, and predictive planning. By factoring in the availability of pilots, drivers, and ships, cuOpt enhances decision-making with real-time insights, reducing transit times, improving resource utilization, and enhancing overall operational efficiency.
Last-Mile Delivery
Efficiently dispatching truck fleets from distribution centers to retail stores and end customers is critical for minimizing costs and meeting delivery expectations. NVIDIA cuOpt optimizes route planning in real time, reducing miles driven, cutting delivery time, and lowering fuel consumption—ultimately decreasing operational costs and reducing pollution for more sustainable last-mile logistics.
Field Dispatch
Effective field dispatch ensures service providers complete scheduled tasks efficiently while accounting for varying job durations and logistical challenges. For example, a telecommunications technician may need to install a router at one location and set up a data cable at another—each requiring different tools, time, and travel routes.
NVIDIA cuOpt optimizes route planning and scheduling, ensuring technicians are fully prepared before departure and follow the most efficient route. This minimizes travel time, maximizes productivity, and enhances service quality, leading to improved customer satisfaction.
Job Scheduling Optimization
Job scheduling is the process of assigning tasks or jobs to available resources—such as machines, workers, or networks—over time to optimize a specific objective, such as minimizing costs and delays, or maximizing efficiency and throughput.
With GPU acceleration, NVIDIA cuOpt enables businesses to make data-driven scheduling decisions, improving operational efficiency and responsiveness in fast-changing environments.
Portfolio Optimization
Effective stock allocation in finance requires strategically distributed investment capital across securities while balancing risk, return, and market dynamics. Investors must navigate volatility, economic indicators, and individual preferences, making real-time adjustments to optimize portfolio performance. The challenge lies in evaluating countless possible combinations and rapidly adapting to shifting market conditions to maintain a competitive edge.
Starting Options
Ways to Get Started With NVIDIA cuOpt
Streamline your optimization problems from data to decisions.
Customer Stories
How Industry Leaders Are Innovating With cuOpt
Lowe’s Transforms Supply Chain Operations With Palantir Ontology and NVIDIA AI
Lowe’s manages its massive supply chain of 7,500 vendors, 130 distribution centers, and 1,700+ stores using AI-powered technology from Palantir and NVIDIA. When disruptions like weather delays occur, intelligent agents use NVIDIA cuOpt to automatically re-optimize shipping routes and allocate resources in real time to maintain seamless operations.
Resources
The Latest in NVIDIA cuOpt Resources
-
Blogs
-
Sessions
-
Training
-
Videos
Next Steps
Ready to Get Started?
Use the right tools and technologies to take logistics optimization projects from development to production.
For Developers
Explore everything you need to start developing with NVIDIA cuOpt, including the latest documentation, tutorials, technical blogs, and more.
Get in Touch
Talk to an NVIDIA product specialist about moving from pilot to production with the security, API stability, and support of NVIDIA AI Enterprise.
FAQs
NVIDIA cuOpt is an open source, GPU-accelerated engine for decision optimization designed to handle large-scale problems with millions of variables and constraints.
cuOpt is designed to excel in mixed-integer programming (MIP), linear programming (LP), vehicle routing problems (VRPs), and quadratic programming (QP).
Yes, NVIDIA cuOpt is an open source engine and is available for developers on platforms like GitHub, PIP, Docker, and Conda.
Mixed-integer programming (MIP) is a type of mathematical optimization where some of the variables are restricted to be integers, while others can be non-integer. MIP is used to model complex optimization problems in areas like resource allocation and scheduling.
Vehicle routing problems (VRPs) are a class of optimization problems focused on determining the optimal set of routes for a fleet of vehicles to service a given set of customers, commonly used in logistics and delivery.
The engine is GPU-accelerated using CUDA capabilities, providing significant speedups over CPU LP solvers when lower-accuracy solutions are acceptable. It is also designed to outperform commercial state-of-the-art VRP solvers.
Yes, enterprise support is available for production deployments through NVIDIA AI Enterprise, which provides security, reliability, and enterprise-class support.
cuOpt agent skills are reusable optimization capabilities that extend the standalone solver into an agentic workflow layer, supporting the full optimization lifecycle from problem formulation to solution interpretation for operations research use cases.
cuOpt is available as open source software on GitHub. It can also be accessed via packaging tools like PIP, Docker, Conda, and NVIDIA NGC.
Yes, developers can explore documentation, tutorials, and technical blogs by visiting the GitHub repository to start developing with cuOpt. Technical blog posts are also available on the NVIDIA Developer Blog.
NVIDIA NIM microservices, which include LLM microservices, can be used to power AI agents to translate natural language business problems into mathematical models and optimized decisions for use cases like supply chain management.
Yes, cuOpt supports dynamic and batch optimization, allowing users to continuously adapt to changing variables and constraints by rerunning models in near real time for optimal decision-making.
In supply chain management, cuOpt-powered AI agents, often integrated with NIM, deliver real-time, optimal resource allocation for greater operational agility, such as pick-up path optimization in warehouses.
cuOpt is integrated with Omniverse digital twins to optimize logistics by simulating real-world fleet operations in a virtual environment, enabling dynamic scheduling, route optimization, and predictive planning for long-haul fleets.
For last-mile delivery, cuOpt optimizes route planning in real time, which reduces miles driven, cuts delivery time, lowers fuel consumption, and ultimately decreases operational costs. One example is its use with Azure Maps for multi-itinerary optimization.
On-demand videos and sessions are available on the NVIDIA On-Demand website. Training materials cover accelerating portfolio optimization and using the route optimization cloud service.
The NVIDIA cuOpt engine is open source and free to use, and cuOpt agent skills are also available at no cost on GitHub. Users can opt for paid enterprise support through NVIDIA AI Enterprise for production deployments.
Yes, you can experience cuOpt immediately for GPU-accelerated decision optimization using Google Colab examples. You can also try an interactive vehicle routing problem example through the NVIDIA API Catalog interface.
Domino’s Pizza
Vehicle Routing at Domino’s: Exploring a GPU-Enabled Approach
Domino’s Pizza delivers thousands of pizzas a day and requires real-time planning and logistics capabilities. Domino’s has implemented a real-time planning system that fits its strict requirements and delivers sub-second runtimes for its use case.
Kawasaki
Reinventing Maintenance Operations With cuOpt and Jetson Orin
Kawasaki Heavy Industries, Ltd. (Kawasaki) is a manufacturing company that’s been building large machinery for more than a hundred years. With NVIDIA cuOpt and NVIDIA Jetson™ Orin, Kawasaki partnered with Slalom, Inc. to transform its track maintenance and inspection capabilities.
Shell
Shell Optimizes Energy Markets With AI-Powered Simulations
Shell is integrating NVIDIA cuOpt to batch optimizations across multiple simulations, addressing unpredictable control issues in power and gas markets. Now, Shell has the ability to make competitive bids, while simultaneously reducing costs and improving efficiency in energy infrastructure; this innovation supports the global transition to lower-carbon energy.
AMPL
AMPL Accelerates Electricity Market Optimization With NVIDIA cuOpt
AMPL, an industry-leading modeling system built specifically for large-scale optimization, has seamlessly integrated NVIDIA cuOpt to revolutionize optimization. By harnessing GPU-accelerated computing, AMPL has reduced problem-solving times from 2 minutes to just 2-3 seconds, significantly enhancing efficiency, scalability, and real-time decision-making in performance-critical energy applications.
Blue Yonder
Blue Yonder Accelerates Last-Mile Delivery With cuOpt
Blue Yonder is transforming supply chain planning and management with AI-powered solutions. Accelerated by NVIDIA cuOpt, its end-to-end supply chain solution platform optimizes last-mile delivery, enabling thousands of deliveries daily across hundreds of vehicles with greater efficiency.
Deloitte
Deloitte Unveils Suite of AI Service Offerings Built on NVIDIA Platforms
Deloitte Compass AI, powered by NVIDIA cuOpt, revolutionizes fleet routing and dispatch optimization by embedding AI directly into workflows. By rapidly processing data and simulating scenarios in seconds, Compass AI enables organizations to reduce costs, improve delivery speed, and enhance customer satisfaction. Announced alongside Deloitte recognition as NVIDIA Consulting Partner of the Year, Compass AI exemplifies cutting-edge AI-driven logistics solutions.
EY
Revolutionize Supply Chain Analytics With AI and Accelerated Computing
EY Supply Chain & Operations Platform (SC&OP) leverages NVIDIA cuOpt and NIM-powered AI agents to optimize supply chain decision-making in seconds instead of hours. By integrating advanced solvers (Heuristics, MIP, LP), SC&OP orchestrates logistics, manufacturing, and sourcing with AI-driven precision—boosting efficiency, agility, and resilience while driving measurable EBITDA improvements.
Lyric
AI-Driven Decision Intelligence for the Modern Supply Chain
Lyric is the first enterprise AI platform for supply chain decision intelligence, empowering organizations to design, plan, and operate with unprecedented agility. By integrating data, algorithms, workflows, and user experiences, Lyric delivers dynamic, AI-driven solutions.
As an NVIDIA partner, Lyric leverages NVIDIA cuOpt to accelerate distribution optimization for major consumer packaged goods (CPG) companies, cutting routing times from 4 hours to just 2 minutes—a 120X speedup—while improving solution quality by 200 basis points. Lyric and NVIDIA cuOpt continue to expand GPU acceleration in supply chain decision-making, boosting efficiency, resilience, and competitive edge at scale.
Microsoft
Enhancing Logistics With Azure Maps and NVIDIA cuOpt for Multi-Itinerary Optimization
At its core, effective route optimization requires reliable inputs to estimate travel times and the ability to apply key constraints such as driver availability, service duration, operating hours, demand, and capacity. In the solution, Azure Maps supplies the essential routing data, while NVIDIA cuOpt processes these constraints to deliver optimized, real-time scheduling and logistics efficiency.
SimpleRose
Turbocharging Prescriptive Analytics and Optimization With GPU Acceleration
As industries face increasing demands in logistics, scheduling, and portfolio management, traditional optimization approaches struggle to scale efficiently. SimpleRose integrates NVIDIA cuOpt to accelerate linear programming (LP) and mixed-integer linear programming (MILP), delivering significant speedups without compromising accuracy.
Slalom
Reinventing Maintenance Operations With cuOpt and Jetson Orin
Slalom, Inc. partnered with Kawasaki Heavy Industries, Ltd. to transform track maintenance and inspection using NVIDIA cuOpt and NVIDIA Jetson™ Orin. With over a century of experience in manufacturing large machinery, Kawasaki leveraged Slalom’s experience to enhance operational efficiency in maintenance processes.