Quantinuum Demonstrates 99.9995% Logical Memory Fidelity on Helios Trapped-Ion Processor

In a technical release and research preprint, full-stack quantum computing provider Quantinuum (NASDAQ: QNT) reported quantum error correction (QEC) performance metrics on its commercial Helios trapped-ion quantum computer, measuring a logical memory fidelity of 99.9995%—equivalent to a logical error rate of 5 × 10^-6 per syndrome-extraction round on a distance-7 code.

Detailed in the technical report ("Quantum error correction well below the color code threshold"), the results evaluate triangular color codes across code distances d = 3, d = 5, and d = 7 on the 137-barium (¹³⁷Ba⁺) QCCD processor. The measured logical error rates demonstrate exponential error suppression as code distance increases, following an experimental suppression factor of Λ = 12.

Key Architectural & Hardware Observations

The study evaluates three core implementation factors within the full-stack architecture:

• Exponential Distance Scaling (Λ = 12): Each step increase in code distance suppressed logical memory errors by a factor of 12. Based on this measured scaling parameter, Quantinuum projects that target logical error rates of 10^-12 across 100 logical qubits can be achieved with ~30,000 physical qubits under this architecture, compared to higher qubit counts required under lower Λ regimes.

• Dual-Layer Leakage Reduction: To mitigate state leakage out of the computational clock states into non-computational levels, Helios incorporates 3-cycle mid-circuit 1762 nm optical repumping at the hardware level (reducing leakage by 6.5x) alongside software-driven SWAP-superdense leakage reduction circuits that cycle data and ancilla qubits through QCCD ion transport.

• Sub-Millisecond Neural Network Decoders: In addition to the A^*-search Tesseract decoder, Quantinuum trained 3D Convolutional Neural Networks (3D CNN) and Round-Token Transformers (RTT) on emulator data and evaluated them on physical hardware without fine-tuning. Both neural models achieved per-round latencies between 0.2 ms and 0.5 ms on NVIDIA H100 hardware, demonstrating compatibility with real-time streaming decoding.

Distinguishing Logical Computation vs. Magic-State Cultivation Metrics

Following the benchmark release, Quantinuum published a technical analysis contextualizing its 99.9995% logical memory fidelity against recent industry milestones, specifically addressing Google Quantum AI's magic-state cultivation results on the Willow processor.

Quantinuum noted that while Google's recent work achieved a 99.99% magic-state fidelity using a cultivation protocol on Willow, the two results measure distinct aspects of the fault-tolerant stack: Google's experiment measured the preparation fidelity of a non-Clifford resource state using post-selection (retaining ~8% of attempts), whereas Quantinuum's Helios benchmark measured the end-to-end logical memory accuracy consuming these operations. In direct comparison on magic-state preparation, Quantinuum previously demonstrated 99.993% magic-state fidelity with an 85% retention rate in 2025 using trapped-ion QCCD architecture, requiring significantly smaller space-time overheads.

The empirical results align with the performance targets established for Apollo, Quantinuum's planned universal fault-tolerant quantum system targeted for 2029.

Review the corporate announcement on Quantinuum Press here, examine the technical breakdown on Quantinuum Memory Blog here, read the comparative analysis on Quantinuum Comparison Blog here, inspect the technical manuscript on Quantinuum Portal here, read QCR's coverage of Google Quantum AI's magic-state cultivation here, and examine QCR's analysis of Google & DeepMind's modular hyperbolic codes here.

October 19, 2026 / Updated: October 22, 2026