November 16, 2025 -
Building on our recent breakthrough in quantum circuit compression, we've extended our AI-powered compression capabilities to support additional native gate sets used by leading quantum hardware providers. Once again, we have achieved record-level performance, and are demonstrating this by sharing our results for these additional gate sets. By advancing circuit compression, we’re bringing useful quantum computations closer to reality.

As today’s quantum hardware, NISQ systems, continue to struggle with lengthy computations because of noise and limited coherence times, hardware-specific circuit compression becomes even more critical.

Last month, QMill demonstrated record-breaking performance with IBM gate sets — reducing, for example, the MOD5_4 benchmark circuit to just 24 gates compared to Quarl's 50 gates; in other words the number of gates was cut in half. An important principle of this service is that our method preserves the full functionality of the circuit, meaning the unitary is preserved in every modification.

Now, our compression method has been expanded to support:

- Ion Trap gate sets (native to IonQ and other trapped-ion systems)
- Rigetti gate sets (native to Rigetti's superconducting quantum processors)

By supporting multiple gate sets, our AI-powered compression method aligns to each hardware platform's native gate set, finding optimal compression strategies that preserve functionality while minimizing gate count and circuit depth. QMill is adding more hardware providers and gate sets on a continuous basis.

Demonstrating our results across gate sets

Like last month, we're demonstrating the capabilities of our method by sharing benchmark results across these new platforms:

Rigetti Gate Set

• For the GF2^8_MULT benchmark circuit, our method compressed the original 4,213 gates down to just 1,597 gates in only 1 hour of runtime — outperforming QUESO, which achieved 1,739 gates after 24 hours of computation. Our method not only yielded fewer gates but also required over 95% less runtime.

Ion Trap Gate Set

• The qcla_mod_7 circuit was reduced from 2,494 gates to 1,099 gates — a 56% reduction, surpassing GUOQ’s result of 1,324 gates (47% reduction). Similarly, the barenco_tof_3 benchmark was compressed from 160 gates down to just 65 gates — a 59% reduction, substantially outperforming GUOQ’s compression to 87 gates (46% reduction).

At QMill, we continue to develop quantum-advantage algorithms for NISQ computing and to improve the utilized techniques to maximize potential of current and near-future quantum hardware.

* * *

Read the previous blog (Oct 15, 2025): QMill sets a world record in quantum circuit compression

* * *

About the author: Clayton Leite has a background in artificial intelligence and focuses on developing cutting-edge AI-driven methodologies for quantum circuit design and optimization.