What’s New in Gurobi 13.0?

Gurobi 13.0 delivers faster performance and introduces major innovations: 

  • 16% faster on MIP, over 2× faster on MINLP
  • New PDHG algorithm implementation for very large LPs, including GPU support (beta feature)
  • Nonlinear barrier method (preview feature) for locally optimal solutions
  • Enhanced tools for heuristics, callbacks, solution pool, and Kubernetes autoscaling  

What kind of performance improvements can I expect?

Gurobi 13.0 delivers, on average, ~16% faster solves on difficult MIP models (>100s) and over 2× faster solves on non-trivial MINLP models (>1s). 

What is the Primal-Dual Hybrid Gradient (PDHG) algorithm?

PDHG is an alternative method for solving extremely large LPs, especially useful when a low-accuracy solution is sufficient for the optimization application.

When should I use PDHG vs. Simplex or Barrier?

Use PDHG when models are too large for Simplex or Barrier to solve efficiently, or when approximate solutions are acceptable. 

Can I run PDHG on CPUs and GPUs today?

Yes, our PDHG implementation is designed to run on both CPUs and GPUs, but please note that functionality for GPUs is still in the beta stage.

What is the nonlinear barrier method?

The nonlinear barrier method is an algorithm for computing locally optimal solutions for continuous nonlinear models. 

What does it mean to compute a “locally optimal solution”?

A locally optimal solution is one such that no better solution exists nearby. While not guaranteed to be the global optimum, it can be much faster to obtain and is often sufficient for many applications. 

What does “preview” or “beta” feature mean?

Preview feature: These features are fully supported and functional in Gurobi 13.0, but may change significantly (including breaking changes in API, behavior or packaging) in future Gurobi technical or major releases.  

Beta feature: These features are not yet intended for production use. Support is provided on a best-effort basis.   

What’s new in heuristics?

Gurobi 13.0 improves the NoRel heuristic by allowing it to use variable hints, and by adding a new NoRelHeurSolutions parameter to stop the heuristic once a set number of solutions is found. This helps users obtain good starting solutions faster on hard MIP models. 

What’s new in callback controls?

Users can now restrict callbacks to specific events (where values). This reduces unnecessary communication—especially on Compute Server—and can improve performance by more than 2× in models that generate many solutions during the solve. 

What new functions are supported in nonlinear models?

Gurobi 13.0 adds operation codes for the hyperbolic tangent and signed power functions, making our nonlinear API more expressive. 

How do the new solution pool quality attributes help me?

The new solution pool quality attributes allow simple assessment of constraint/bound/integrality violations across all solutions, helping compare solution quality tradeoffs. 

 

How do the new multi-objective attributes improve workflow?

Objective values and MIP gaps for all the individual objectives can now be queried after solving. 

Can I limit the time spent on MIP starts?

Yes. New parameters StartTimeLimit and StartWorkLimit let you control how much time or work units are spent at most processing MIP starts—helpful if you’re under strict solve time limits. 

Do I need a separate license for these features?

No. All new features and improvements in Gurobi 13.0 are included with your standard license. 

Notes: 

  • To run PDHG on GPUs (not available on Instant Cloud or Compute Server), download the GPU-enabled binary from the Gurobi Downloads page and run it on your own hardware. 
  • The Kubernetes Autoscaler Operator is designed for use with WLS Compute Server licenses and will be available on our DockerHub page. 

Do I need special hardware (GPU) to benefit?

Yes, a modern NVIDIA GPU is required to take advantage of the new PDHG on GPU capability in Gurobi 13.0. If you don’t currently have access to GPUs but want to understand whether such hardware could benefit you, you can submit your models to Gurobi for evaluation.

Are there any accessibility improvements in this release?

Yes. Gurobi 13.0 upgrades Cluster Manager accessibility, aligning with WCAG 2.1 AA standards. Enhancements include clearer heading structures, improved contrast, and streamlined navigation, making it easier for allusers to access features consistently across web platforms. 

Will PDHG on GPUs and nonlinear barrier be stable and supported in the long term?

Yes. Both features are fully supported today and form part of our long-term roadmap. 

How will Gurobi continue expanding into nonlinear optimization and GPU acceleration?

Gurobi will continue investing in R&D to expand our nonlinear optimization and accelerated computing capabilities, with a focus on delivering clear, practical benefits for our customers. 

Additional Insight

Guidance for Your Journey

30 Day Free Trial for Commercial Users

Start solving your most complex challenges, with the world's fastest, most feature-rich solver.

Always Free for Academics

We make it easy for students, faculty, and researchers to work with mathematical optimization.

Try Gurobi for Free

Choose the evaluation license that fits you best, and start working with our Expert Team for technical guidance and support.

Evaluation License
Get a free, full-featured license of the Gurobi Optimizer to experience the performance, support, benchmarking and tuning services we provide as part of our product offering.
Cloud Trial

Request free trial hours, so you can see how quickly and easily a model can be solved on the cloud.

Academic License
Gurobi provides free, full-featured licenses for coursework, teaching, and research at degree-granting academic institutions. Academics can receive guidance and support through our Community Forum.

Search

Gurobi Optimization

Navigation Menu