Question about the hybrid CQM Solver

I'm currently trying the hybrid constrained quadratic model (CQM) solver and have some questions.

1. Is it guaranteed that we can always obtain some feasible samples (at least one sample) for any problem? As far as I tried, I always obtained some feasible samples.

2. Is the "charge_time" of the hybrid CQM solver is always 5,000,000 microsecond? Although I tried some problems with different sizes, the obtained "charge_time" was always the same value.

3. Is the price of the hybrid CQM solver per "charge_time" the same as that of the hybrid BQM solver?



  • Q2 has been sorted out. The "minimum_time_limit" of the CQM is determined as the larger one of the following:

    A:1.57e-04 x (# of variables) + 4.65e-06 x (# of biased) + 6.44e-09 x (# of variables) x (# of constraints),

    B: the value of "minimum_time_limits" (5 by default).

    It seems that we cannot set the value of "minimum_time_limits" less than 5 at this time.

    Comment actions Permalink
  • Hi Daigo,

    Hope you are finding the LeapHybridCQMSampler useful!

    Here are the answers to your questions:

    1. No, a feasible solution is not always guaranteed. Some problems might have constraints that are impossible to meet, or the others might be hard enough for which the CQM solver will fail to find a feasible solution in the specified time.

    2. Looks like you already found the answer to this question! As you mentioned, the minimum run time for the hybrid CQM solver is 5 seconds. For more complex problems you might need a longer run time which can be set using the time_limit parameter(see the docs here:

    3. You can find the rate at which a particular solver consumes access time in the solver’s quota_conversion_rate property. If you have any further questions regarding our plans and pricing, please contact our sales team at

    Comment actions Permalink
  • Hi Tanvi,

    I understood your explanation fully, thank you!

    I feel the LeapHybridCQMSampler is really useful. The BQM sampler returns only one sample, but the CQM sampler returns multiple samples in one shot. Furthermore, the CQM sampler tends to be faster than the BQM sampler for large size constrained problems as far as I tried.

    Comment actions Permalink

Please sign in to leave a comment.

Didn't find what you were looking for?

New post