What's coming
Our public roadmap reflects current priorities — subject to change as we learn from users. Have a feature request? Let us know.
General availability of 8-GPU parallel annealing with automatic load balancing and fault tolerance. Extends problem size ceiling to ~40,000 variables.
Production-ready on-premise container with SSO integration, RBAC, and complete audit logging for regulated industries.
Submit incremental variable updates to a running solver job — enabling real-time re-optimization as constraints change.
Community-submitted instances, automated nightly leaderboard refreshes, and solver comparison permalinks for sharing results.
A new solver category using matrix product states (MPS) — particularly strong on structured constraint satisfaction problems and near-planar graphs.
Connect QAOA and VQE to real quantum processors via IonQ and IBM Quantum Cloud. Automatic fallback to simulation when hardware queue exceeds threshold.
Describe your objective and constraints in Python and let AutoQUBO derive the QUBO matrix automatically. No manual penalty tuning required.
Shared experiment workspaces where teams can co-own runs, annotate results, and track solution quality over time across projects.
Reserved GPU cluster plans for enterprise customers — dedicated A100 nodes with guaranteed throughput SLAs and no shared queue.
Bridge between CP-SAT (OR-Tools) and NEROX — automatically routes problems to the best solver based on structure, with fallback.
For each solution, the explainability layer surfaces which variables changed most between runs, energy landscape visualization, and constraint violation heatmaps.
Long-running solver jobs that accept real-time constraint updates and emit improved solutions as they arrive — for dynamic optimization applications.
As coherent photonic processors reach commercial availability, NEROX will provide a unified interface across photonic, quantum, and GPU solvers.
Intel Loihi and BrainScaleS backend support for energy-efficient combinatorial optimization at the edge.
Machine learning model that predicts the optimal solver and hyperparameter configuration from problem structure — zero configuration required.
GPU-native quantum-inspired optimization for combinatorial problems at any scale.