NVIDIA/cuopt v26.04.00

v26.04.00

Repository: NVIDIA/cuopt

Tag: v26.04.00

Published: 2026-04-09T20:42:17Z

Prerelease: no

Release notes:

What's Changed

🚨 Breaking Changes

• Update solution object to include the LP solver by @nguidotti in https://github.com/NVIDIA/cuopt/pull/822

🐛 Bug Fixes

• Fix doc version update script to handle new url by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/810
• Add nvjitlink to cuda-toolkit pip extras for cusparse compatibility by @bdice in https://github.com/NVIDIA/cuopt/pull/850
• Fix issue with incorrect lower bound on air05 by @chris-maes in https://github.com/NVIDIA/cuopt/pull/890
• Fix batch pdlp and python api support by @Kh4ster in https://github.com/NVIDIA/cuopt/pull/893
• Fix timer checks by @akifcorduk in https://github.com/NVIDIA/cuopt/pull/849
• fix how tests create service subprocess (causes errors with new cuda) by @tmckayus in https://github.com/NVIDIA/cuopt/pull/897
• Fix thrust tuple includes + more timer checks by @aliceb-nv in https://github.com/NVIDIA/cuopt/pull/902
• Fix race condition in add_external_solutions_to_population by @akifcorduk in https://github.com/NVIDIA/cuopt/pull/909
• Update to PSLP version with bug fixes by @rg20 in https://github.com/NVIDIA/cuopt/pull/920
• Fix variable bound violation in CPUFJ moves by @aliceb-nv in https://github.com/NVIDIA/cuopt/pull/930
• Fix #903: Validate unique vehicle_ids in VRP fleet data by @anandhkb in https://github.com/NVIDIA/cuopt/pull/917
• Fix incorrect errors with x + x*x, +x, -x in Python API by @chris-maes in https://github.com/NVIDIA/cuopt/pull/936
• turn off cpu only tests to unblock CI during investigation by @tmckayus in https://github.com/NVIDIA/cuopt/pull/959
• Fix lexicographic search by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/969
• apply -UNDEBUG to coupt flags instead of global settings by @tmckayus in https://github.com/NVIDIA/cuopt/pull/964
• papilo probing workaround, probing cache sync, extra error logging by @aliceb-nv in https://github.com/NVIDIA/cuopt/pull/992
• Add exception handling for pdlp in concurrent mode by @Iroy30 in https://github.com/NVIDIA/cuopt/pull/966
• Implement UnboundedOrInfeasible termination status by @rg20 in https://github.com/NVIDIA/cuopt/pull/941
• Fix race in graph capture when set_simplex_solution() is called by @aliceb-nv in https://github.com/NVIDIA/cuopt/pull/1017
• Fix issue with infinite lower bounds and try to bound free variables in barrier by @chris-maes in https://github.com/NVIDIA/cuopt/pull/1001

📖 Documentation

• Add install selector UI by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/961
• update doc for usage of incumbent utility function by @Iroy30 in https://github.com/NVIDIA/cuopt/pull/947
• Disable running notebooks while building docs by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/980
• Fixing broken links by @cwilkinson76 in https://github.com/NVIDIA/cuopt/pull/1028
• Doc and release notes updates by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/1006

🚀 New Features

• Add memory model support for host-resident problem data by @tmckayus in https://github.com/NVIDIA/cuopt/pull/819
• Add support for FP32 and mixed precision in PDLP by @Kh4ster in https://github.com/NVIDIA/cuopt/pull/910
• Remote execution implementation by @tmckayus in https://github.com/NVIDIA/cuopt/pull/939
• Add cuopt_grpc_server as an option for the cuopt image by @tmckayus in https://github.com/NVIDIA/cuopt/pull/1003

🛠️ Improvements

• fix(build): build package on merge to release/* branch by @gforsyth in https://github.com/NVIDIA/cuopt/pull/800
• Main merge release/26.02 by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/801
• Main merge release/26.02.update by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/812
• move ruff configuration into pyproject.toml by @jameslamb in https://github.com/NVIDIA/cuopt/pull/806
• refactor: drop support for Python 3.10 by @gforsyth in https://github.com/NVIDIA/cuopt/pull/809
• tighten wheel size limits, expand CI-skipping logic, other small build changes by @jameslamb in https://github.com/NVIDIA/cuopt/pull/804
• remove pip.conf migration code in CI scripts, fix pr-builder configuration by @jameslamb in https://github.com/NVIDIA/cuopt/pull/821
• Use verify-hardcoded-version pre-commit hook by @KyleFromNVIDIA in https://github.com/NVIDIA/cuopt/pull/823
• Refactoring agentic skill files based on roles and modules by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/818
• Main merge release/26.02 4 by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/830
• Rename and split up cpp/src directories by @chris-maes in https://github.com/NVIDIA/cuopt/pull/846
• Consider objective function integrality when pruning + bug fixes by @aliceb-nv in https://github.com/NVIDIA/cuopt/pull/851
• Simplex work estimates by @chris-maes in https://github.com/NVIDIA/cuopt/pull/856
• Update raft headers by @aamijar in https://github.com/NVIDIA/cuopt/pull/865
• Disable release/26.02 nightly and enable release/26.04 nightly by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/895
• Add .github/release.yml for auto-generated release notes by @AyodeAwe in https://github.com/NVIDIA/cuopt/pull/898
• Use sccache-dist build cluster for conda and wheel builds by @trxcllnt in https://github.com/NVIDIA/cuopt/pull/848
• Type and execution path cleanup for memory model changes by @tmckayus in https://github.com/NVIDIA/cuopt/pull/908
• Disable auto summary from code rabbit by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/913
• Fix routing module bugs from issue #857 by @anandhkb in https://github.com/NVIDIA/cuopt/pull/914
• Upgrade skills by @rgsl888prabhu in https://github.com/NVIDIA/cuopt/pull/918
• Clique Table Generation by @akifcorduk in https://github.com/NVIDIA/cuopt/pull/627
• update exceptions and guards by @Iroy30 in https://github.com/NVIDIA/cuopt/pull/912
• Add clique cuts by @akifcorduk in https://github.com/NVIDIA/cuopt/pull/937
• Fix MIR cut aggregation. Complement integer variables to find MIR cuts. Use Variable Lower/Upper Bounds in MIR cuts. by @chris-maes in https://github.com/NVIDIA/cuopt/pull/905
• Add MIP gap to heuristic log during dual-simplex root relaxation by @anandhkb in https://github.com/NVIDIA/cuopt/pull/942
• feat: add support for Python 3.14 by @gforsyth in https://github.com/NVIDIA/cuopt/pull/933
• update routing tests by @Iroy30 in https://github.com/NVIDIA/cuopt/pull/940
• Add a skill for evolving skills or…