Hi
No, you can't.
CP is intrinsically sequential. Some parts can be parallelized, but it seems that the best is to solve the same problem with different search strategies on different threads.
This can be done with Choco-solver through ParallelPortfolio.
Nothing is done to exploit GPUs, at least in choco-solver. I suppose that a way would be to use GPU to propagate constraints, but the bottleneck would be the intersection of resulting domains...
Best regards
CP