Catalog — Comparison
Compare quantum processors
Two processors, side by side, with the sources and measurement methods attached — and a mandatory statement of why these numbers may not be directly comparable. QPU137 never declares a universal winner: different architectures are good at different circuits.
Pick two gate-model devicesnon-gate models are excluded — counts aren’t comparable
ibm_boston (Heron r3) vs Quantinuum System Model H2 (H2-1 / H2-2)retrieved 2026-08-19
| Metric | ibm_boston (Heron r3) | Quantinuum System Model H2 (H2-1 / H2-2) |
|---|---|---|
| Technology | Superconducting | Trapped ion |
| Execution model | gate-based | gate-based |
| Topology | heavy-hex | all-to-all |
| Qubit count | 156 vendor claim | 56 vendor claim |
| Two-qubit gate quality | 6.43E-4 vendor claimbest (lowest single edge) | 1e-3 (Typical), 2e-3 (Max) vendor claimTypical = most representative infidelity during machine operation, averaged over all operational zones; Max = upper bound during operation |
| Single-qubit gate quality | not publicly disclosed | 3e-5 (Typical), 2e-4 (Max); measured H2-1: 1.9e-5 (+/- 4.2e- vendor claimTypical = average over all operational zones (data sheet); per-machine averages (performance validation); system average (PRX paper) |
| Coherence | not publicly disclosed | T1: 'several minutes' (qualitative); T2: no value published vendor claim |
| Readout | 3.418E-3 vendor claimmedian (column labeled 'Readout error (median)') | not publicly disclosed |
| Gate duration | not publicly disclosed | not publicly disclosed |
Finding — why these numbers may not be directly comparable
- Different technologies — Superconducting and Trapped ion trade connectivity, speed, and coherence very differently; single figures hide the trade.
- Measurement methods differ. Gate-quality figures may be randomized benchmarking, layered benchmarks, or application-level results; medians are not best-pairs. Open each figure’s record on the device pages for its stated method and aggregation.
- Qubit counts rank nothing by themselves. Usable width depends on connectivity and your circuit’s shape — compile your own circuit against both architectures instead.
- Vendor claims are labeled. Most are accurate, but a claim and an independent measurement are different kinds of evidence.
- Snapshots drift. Calibration figures change; both columns were retrieved 2026-08-19.
QPU137 declares no overall winner. A single ranking would hide the trade-off that matters to your circuit.