FirstQFM Reports 33% Faster Quantum Gates Using AI
FirstQFM, a Sweden-based company, has announced that its quantum foundation model successfully reduced single-qubit gate execution time by 33% on a QuantWare superconducting processor without any loss of gate fidelity. The test took place at the Israeli Quantum Computing Center on a 21-qubit QuantWare Contralto-D21 QPU, and utilized Quantum Machines control hardware and NVIDIA accelerated computing.
Indeed, this achievement, announced on October 15, highlights a growing trend in the quantum field: leveraging classical AI models to handle calibration tasks that typically consume valuable quantum processor time.
FirstQFM’s Approach and Findings
Specifically, FirstQFM’s model generated custom pulse-level gate implementations using existing calibration data from the Quantum Machines calibration stack, which eliminates the need for new quantum hardware measurements. The company reported that randomized benchmarking confirmed no drop in fidelity for these shorter gates on the QuantWare chip.
The model also addressed two-qubit gates, which are more challenging to calibrate and are crucial for entanglement. By analyzing existing calibration data and performing a brief hardware probe, FirstQFM identified five qubit pairs exhibiting higher Bell-state fidelity. These improvements were subsequently validated through independent hardware runs.
The key details of the reported results and setup include:
- Processor: QuantWare Contralto-D21 (21-qubit superconducting).
- Single-qubit gate time: 33% reduction compared to the baseline (specific nanosecond durations not disclosed).
- Fidelity: “No loss” as determined by randomized benchmarking (no before/after percentages provided).
- Two-qubit result: Improved Bell-state fidelity for five qubit pairs (specific pair identities and values not stated).
- Control and calibration: Quantum Machines OPX hardware and QUA software.
- Model development: NVIDIA Leonardo supercomputer and the CUDA-Q open-source platform.
Vish Ramakrishnan, CEO and co-founder of FirstQFM, stated that the firm’s “proprietary, patented foundation models are built to learn the operating characteristics of quantum hardware” to generate shorter gates as they maintain fidelity.
Significance of Reduced Gate Times
The duration of single-qubit gates directly impacts coherence budgets. Faster gates allow for more operations within a qubit’s limited coherence time, which is critical for executing deeper circuits in the NISQ (Noisy Intermediate-Scale Quantum) era before noise degrades the signal. Shifting optimization from the QPU to classical computing also conserves valuable machine hours and simplifies calibration as chips scale in qubit count, couplers, and crosstalk.
Calibration drift and gate speed are also crucial factors in the complex challenge of error correction, a topic QNews has previously covered, including work like Alice & Bob’s GPU decoding of LDPC-cat codes. For more updates on quantum hardware and software, continue reading QNews.