QONTOS-1 through QONTOS-5 form a single architectural arc with attenuating confidence. QONTOS-1 is an engineering target. QONTOS-2 and QONTOS-3 are scenarios, committed in engineering shape and conditional in numerical envelope. QONTOS-4 and QONTOS-5 are research-vision architecture studies, not product commitments. No generation carries a calendar date; every generation carries a gating expression.
Each generation is sized against the engineering risk retired by the one before it.
The QONTOS family arc starts with the first-machine assembly and ends with a datacenter-scale research study. Each row below is gated by the acceptance of the one before it. The status keys are exact.
Acceptance gates retire risks; calendar dates follow gates, not the other way round. The QONTOS-1 G1 → G4 sequence is the prototype for every successor. Dates below are engineering targets contingent on the preceding generation's full acceptance.
Each gate retires a class of risk; each generation extends the envelope of the one before it. The dates above are programme targets; their slip behaviour is governed by the gating expression of the preceding generation. There are no skip steps.
Log-log map of physical qubits against logical qubits. Solid markers anchor the engineering target. Outlined markers are committed scenarios. Dashed markers indicate research-vision architecture studies that require multiple breakthroughs landing together.
Module organisation, interconnect regime, decoder, error correction, workload class, and the gating expression that opens engineering scope. Numerical envelopes are committed for the target generation, conditional for scenarios, and explicitly speculative for the research vision.
QONTOS-1 retires architectural risk: two modules of five hundred transmons, connected by a base-regime photonic interconnect at η ≥ 0.1 %, validated end-to-end by an operating software runtime. The first logical qubit at distance d = 5 is the G4 acceptance criterion. The machine does not claim useful fault-tolerant computation or ε_L below p_phys; it claims that the hybrid superconducting-photonic architecture is operable end-to-end and that its named risks retire at measurable gates.
QONTOS-2 raises the surface-code distance from d = 5 to d = 9 → 11 and increases module count from 2 to a band of 4 to 8, anchoring the first sustained logical-qubit advantage on a hybrid superconducting-photonic platform. Modules pair into supermodule couplets sharing a single dilution refrigerator at ≥ 50 μW cooling power at 20 mK. Where QONTOS-1 demonstrates that the architecture is operable, QONTOS-2 demonstrates that operating it delivers a logical error rate strictly below the underlying physical error rate, sustained across continuous workloads at the lattice-cycle cadence.
QONTOS-3 is a scenario for the first useful-work pilot if QONTOS-2 closes its logical-error and purified-pair gates. Sixteen to thirty-two modules are organised in racks of four, with a shared cryoplant per rack and a rack-level photonic switch matrix routing Bell-pair traffic between module pairs within and across racks. The interconnect graduates to a multi-rack photonic mesh with wavelength-division multiplexing. Decoder and distillation factories move from validation plans into system-level engineering assumptions.
QONTOS-4 is a research-vision architecture study rather than an engineering target. It explores what a 64 to 128 module production-class system would require: redundancy, hot-spare modules, fleet-level calibration, purified photonic supply, and bounded-latency decoding beyond the QONTOS-3 scenario. The numbers are planning envelopes, not commitments.
QONTOS-5 is a long-range architecture study, not a committed facility build. A million physical qubits requires roughly 2,000 modules at 500 qubits per module; the lower bound of 500 modules corresponds to roughly 250,000 physical qubits. The v5.4 paper uses this envelope as a resource-estimate closure exercise against external RSA-2048 benchmarks, not as a claim that the programme can beat those benchmarks with fewer qubits today.
The QONTOS programme states each generation against a precise evidence basis. The status key sets the rules under which numerical envelopes are committed, conditional, or speculative.
The engineering shape is committed and the numerical envelopes are anchored in the v5.4 whitepaper. QONTOS-1 is the only generation at this status. Each subsystem has a measurement protocol and a passage gate; the machine is the assembly that proves they hold simultaneously.
The engineering shape is committed; the numerical envelopes are conditional on the prior generation's acceptance criteria having passed. QONTOS-2 and QONTOS-3 sit at this status. The architecture is sized to deliver the scenario when the gating expression closes.
QONTOS-4 and QONTOS-5 sit at this status. Their numbers identify open research constraints such as purified-pair supply, d ≥ 21 decoder fabric, distillation overhead, and facility-scale cryogenics; they are not engineering targets.
Gating discipline. No generation carries a calendar date. Each generation carries a gating expression — a specific set of subsystem milestones that must close before the generation enters engineering scope. The same discipline that governs the four G1 → G4 gates of QONTOS-1 extends across the family arc: each transition is conditional on specific, measurable evidence from the prior generation. Numerical targets are stated as engineering objectives gated by subsystem milestones, not as measured performance.
Eight engineering risk categories, five generations, one cumulative retirement map. A solid cell signals a risk closed at acceptance of that generation. An outlined cell signals a committed reduction. An open cell signals a research item still on the programme register.
The matrix attenuates left to right. Risk categories that QONTOS-1 closes at G4 do not reopen for later generations; later cells inherit the closure. The research categories that remain open through Gen III become architecture-study inputs at Gen IV and Gen V. This is the discipline that makes the family arc a single architectural programme rather than five independent designs.
The QONTOS family is paced by what each generation enables, not what it scales. Numerical thresholds open or close particular classes of workload. Below: the canonical workload class per generation, with representative example targets.
Section 11 of the v5.4 architecture whitepaper anchors the family arc. The QONTOS-5 envelope is treated as a resource-estimate closure exercise against external RSA-2048 benchmarks, not a committed facility claim.
The QONTOS-2 generation will pair two compute modules into a supermodule couplet under a single dilution refrigerator at ≥ 50 μW cooling power. The supermodule footprint is the unit of scaling for QONTOS-3.
The MWPM decoder will graduate from FPGA implementation at QONTOS-1 and QONTOS-2 to dedicated ASIC tiles at QONTOS-3. The QONTOS-4 generation runs the same ASIC pattern at fleet scale with hot-spare redundancy.
The QONTOS programme develops modular superconducting-photonic quantum computers across a five-generation arc. QONTOS-1 is the first-machine assembly that proves the architecture. QONTOS-2 and QONTOS-3 are conditional scenarios. QONTOS-4 and QONTOS-5 remain research-vision architecture studies.
Each generation is sized against the engineering risk retired by the one before it. The first commercially relevant logical-qubit count is not claimed by QONTOS-1; later generations depend on purified photonic links, protected distillation overhead, and decoder scaling that remain open research and engineering topics.
QONTOS-1 specification