GPU Accelerated Vehicle Routing, Managed for you.

Run NVIDIA cuOpt on your data in seconds. No GPUs to procure, no CUDA to configure, no infrastructure to manage. Upload your problem, get optimized routes.

HOW IT WORKS

Three Steps to Optimized Routes.

1. Define Your Problem

Upload your vehicles, sites, tasks, and constraints. Use our structured Excel template, submit via our Python library, or upload a .json file.

2. Solve On Managed GPUs

cuOpt runs on NVIDIA GPUs managed by SimpleRose. No GPU procurement, no CUDA, no Docker, no cloud infrastructure to configure

NVIDIA logo

3. Analyze Results

Interactive solve summary, Gantt charts, per vehicle metrics, and exportable data. See exactly how your fleet is utilized.

0s
to solve a VRP with 10,000 locations on 1 GPU
0x
Speedup over CPU solvers on large LP problems
20
GPUs, drivers, or containers to manage

WHO IT’S FOR

Get Started From Where You Are.

Already Using NVIDIA cuOpt?

Run your existing cuOpt models on managed GPUs. Submit via the Python library and let us handle the infrastructure. Same API, zero DevOps.

Using Another Routing Tool?

Benchmark cuOpt against your current solver on your own data. See if GPU acceleration delivers faster solves and better routes for your problem.

New To VRP Optimization?

Define your problem in our structured Excel template. Vehicles, sites, tasks, time windows, capacities. No coding or solver expertise required.

SUPPORTED VRP FEATURES

Real World Constraints, Not Toy Problems.

Time Windows

Vehicle availability windows and task delivery windows with configurable start and end times.

Multi Capacity

Multiple capacities such as weight, pallets, and volume with heterogeneous fleet support.

Pickup & Delivery

Paired pickup and delivery tasks with precedence constraints and capacity tracking.

Breaks & Shifts

Driver breaks with time windows and allowed locations. Service times per stop.

Cost Optimization

Fixed and variable vehicle costs. Route cost and duration limits. Priority weighting for cost versus time.

Optional Tasks

Tasks that can be skipped if they do not fit. Prize collecting objectives for revenue maximization.

FAQ
Common Questions.

Rose cuOpt is a managed SaaS platform built by SimpleRose for running NVIDIA cuOpt on GPUs. You upload your vehicle routing problem, we handle the GPU infrastructure, and you get optimized routes back — with interactive results including Gantt charts and per-vehicle metrics.

No. Rose cuOpt runs on managed NVIDIA GPUs. You do not need to procure hardware, install CUDA, configure Docker, or manage any cloud infrastructure.

Three ways: Use our structured Excel template, submit via our Python library, or upload a .json file.

Solve times depend on problem size and complexity. A 10,000 location VRP with real world constraints solves in approximately 30 seconds on 1 GPU. For large LP problems, cuOpt achieves up to 5,000x speedup over CPU solvers.

Yes. NVIDIA open sourced cuOpt in 2025 and it won the COIN OR Cup. The solver is free. The barrier to adoption is GPU infrastructure and operational complexity, which Rose cuOpt removes.

Rose cuOpt is currently free during our beta period. In exchange, we ask beta users to share detailed feedback on their experience to help us build the right product.

The Rose cuOpt interface currently supports vehicle routing problems. If you would like to benchmark cuOpt on other problem types such as LP, MILP, or other combinatorial optimization, contact us and we will set it up.

Not yet. The Rose parallel MIP solver is coming soon as a separate product. Rose cuOpt currently focuses on NVIDIA cuOpt for GPU-accelerated vehicle routing. Stay tuned for updates on Rose solver availability.

FROM OUR BLOG

Learn More About GPU Optimization.

Try Rose cuOpt

on your data.

Rose cuOpt is free during beta. All we ask is that you share your feedback on the experience so we can keep making it better.