SoftBank used quantum computing to group deliveries. A classical planner chose the routes
Across 1,921 Tokyo destinations, the hybrid approach produced 2.31% less total distance than grouping and routing with classical methods.
Reading tools for text size, copying, printing and your reading list

Grouping delivery stops before planning each route shortened the total distance by 2.31% in a test on 1,921 destinations across Tokyo’s 23 wards. The comparison used classical computing for both stages. Classiq described the study with SoftBank on November 17.[1]
Where the quantum step fits
Imagine sorting delivery addresses into neighbourhoods before deciding which street to visit first. Hybrid quantum and classical machine learning did the grouping. An ordinary computer still chose the routes within those areas. The reported distance comes from the study’s route plans.[1]
The release does not identify that processor or the baseline algorithms, disclose statistical variation, or compare runtime and cost. The percentage therefore does not establish an advantage over the best classical methods or savings from operational deliveries.[1]
A separate design explores quantum-search routing for future fault-tolerant machines. That is a proposed workflow with resource estimates, not the demonstrated routing stage.[1]