The gap between quantum simulation and real-world quantum computing has always been defined by three things: scale, noise, and the ability to optimize. With this latest MIMIQ™ release, we’ve addressed all three — and added first-class support for the variational algorithms increasingly central to near-term quantum research.
Simulating noise the right way
Real quantum hardware is imperfect, and any simulation that ignores that fact has limited practical value. MIMIQ has supported noise channels for some time, but this release introduces a more principled approach: noise models.
Rather than adding noise channel by channel to individual circuits, users can now define a noise model independently — describing the error profile of a specific quantum computer — and apply it automatically across any circuit. This separation of concerns makes it far easier to benchmark algorithms fairly, compare hardware configurations, and build reproducible workflows.
For teams doing serious hardware-oriented research, this is a meaningful shift in how noise simulation can be structured.
Variational algorithms as a first-class workflow
Variational Quantum Algorithms — particularly VQE — have become a focal point of near-term quantum computing. Until now, running a full VQA workflow required assembling pieces from different tools. MIMIQ now integrates parameter optimization directly into its execution engine.
Users can define parameterized circuits, configure an optimization loop, and offload the computation to MIMIQ’s remote executors — all within a single, coherent workflow. This reduces friction significantly for researchers working on quantum chemistry, optimization problems, or any task that relies on iterative circuit evaluation.
Smarter circuits before execution
Simulation performance is not only about the engine — it also depends on the quality of the circuit going in. Two new optimization features address this directly.
Automatic SWAP gate removal (remove_swaps=true) identifies and eliminates redundant operations that accumulate during compilation, reducing circuit depth without changing the computation. Improved qubit reordering heuristics (reorderqubits) find better qubit orderings for MPS simulations, leading to measurable performance gains and more consistent results across runs.
These improvements compound over time: every unnecessary gate removed and every better ordering found translates into faster, more reliable simulations.
Breaking the memory barrier
Large-scale quantum circuits have long been constrained by available RAM. With this release, MIMIQ introduces MPO streaming — a mechanism that streams Matrix Product Operators during execution rather than holding them entirely in memory.
The result is that circuits which previously exceeded hardware limits can now run without modification. Enabling it requires a single parameter: streaming=true in the execute() call. For researchers working at the boundary of what simulation can handle, this removes what was often a hard ceiling.
Getting started
To access all new features, update your MIMIQ packages:
Python
pip install --upgrade mimiqcircuits-pythonJulia
using Pkg
Pkg.update("MimiqCircuits")Full documentation is available at docs.qperfect.io. We welcome feedback from the community as we continue to develop the platform.