Quantum-inspired annealing for QUBOs at any scale.
A classical heuristic that draws on quantum-annealing intuition — stochastic tunnelling out of local minima, then gradual cooling toward a low-energy state — without the topology constraints of physical qubits.
Feed it fully connected QUBOs with thousands of variables and get back a ranked portfolio of near-optimal samples, ready for scheduling, logistics, portfolio, or network-design pipelines.
Handles fully connected, high-dimensional QUBOs that exact solvers and current quantum hardware cannot reach in a useful time budget.
Converges to near-optimal solutions inside operational windows — fast enough to rerun as inputs change through the day.
Classical infrastructure, quantum-inspired search — packaged for real production timelines.
Adaptive cooling schedule
Temperature drops are tuned to the structure of the QUBO so the sampler explores aggressively early on, then locks in on low-energy states without overshooting.
Incremental delta scoring
Flipping a variable updates the objective by a closed-form delta rather than recomputing it — the inner loop stays tight on dense, large problems.
Parallel annealing runs
Independent trajectories run in parallel across cores and GPUs, raising the odds of reaching the best basin within an operational time budget.
Boltzmann multi-sample output
Return a ranked portfolio of near-optimal solutions rather than a single answer — pick the option that best fits constraints outside the cost function.
Topology-free input
No qubit graph, no embedding step. Submit fully connected QUBOs with thousands of variables exactly as you formulate them.
Sampler-style SDK
Drop-in compatibility with the sampler interface familiar from Ocean and similar frameworks — define the problem, call the API, iterate.
How a Simulated Annealer run produces its sample set.
- 01Formulate the QUBO with any topology you need
- 02Submit through the sampler-style cloud API
- 03Initialise a high-temperature random state
- 04Anneal with stochastic acceptance of uphill moves
- 05Cool on an adaptive schedule toward convergence
- 06Return a ranked portfolio of near-optimal samples