BEIT Logo

QUBO Reference Solver

Inspired by the D-Wave machines, we offer provably correct solution for all QUBO instances restricted to a specific topology. Within seconds, we can provide you with optimal solution for your problem or multiple samples from Boltzmann distribution.

Caveats: not all problems embed nicely onto the supported topologies. If that's the case, consider the following products.

TopologySizeSolutionTime
Chimera½ of the deviceoptimal~1s
Pegasus~½ of the deviceoptimal~1s
Zephyr~½ of the deviceoptimal~1s
Establish a problem model
Map the problem to QUBO
Embed the QUBO onto the topology
Use the Reference Solver

QUBO Annealer

Simulated annealing solution, executed on classical hardware, tuned specifically for tackling QUBO problems. Not limited by the topology and allowing full connectivity between variables for efficient problem representation. Based on the QPLIB benchmark, it wins in most of the cases against commercial general-purpose optimizers, and also the majority of cases against other QUBO- specialized solvers.

Caveats: not all problems can be effectively mapped to QUBO. Particular problem may require too many variables to be feasible for solving. If that's the case, consider the following product.

Evaluation on the based on the QPLIB benchmark. Gurobi & QUBOWL results by H. Mittelmann
QPLIB IDGUROBIQUBOWLOur Solver
350616s~0s~0s
3642299s24s~0s
3650204s11s~0s
36933600s1648s6s
370666s~0s~0s
373812s1s~0s
3822148s~0s~0s
3832297s2s~0s
38383600s3s~0s
38501363s11s61s
387783s~0s~0s
57256s1s~0s
5755~0s~0s145s
58819s18s~0s
58823600s~0s~0s
59092s~0s~0s
Establish a problem model
Map the problem to QUBO
Use the QUBO Annealer

Generic Annealer

The most powerful, heuristic-based solution against all types of problems. Uses techniques from previous products, translated into generalized solutions.

Caveats: whilst being the most powerful technology here, it is not ready to use straight away and requires our efforts to adapt your problem to the engine.

A comparison with 2QAN, another heuristic-based quantum compiler.
PhaseOur compiler2QAN
Placing92.3% success rate within 27 ms~16 minutes
Routing~19,4 swaps30 swaps
Scheduling~7,6 depth~12,5
Establish a problem model
Develop the right expression of the problem in our framework
Use the dedicated annealer