Open-instance verified benchmarks (Heron 156Q)

Five reproducible QPC-orchestrated workloads on the IBM Open plan (open-instance): labelled context layouts from 4 registers on one 156-qubit chip through 24 contexts across six cluster jobs (tier T4). Thirteen IBM Runtime jobs, 512 shots each, schema qpc_benchmark_v1.

What this page is — and is not

This page is orchestration and hardware-reach evidence. Real layouts, real backends, verifiable job IDs, and standard JSON payloads. It shows QPC can plan and execute multi-context workloads on Heron 156Q at increasing width and cluster depth.

It is not, by itself, a proof that polycontextural coupling outperforms a null model. High shot diversity (entropy, unique outcomes) is expected for deep circuits on good hardware — including random circuits — and does not uniquely signature QPC. Controlled ablations and physical observables on other site pages carry that stronger claim.

Summary

Audit: Every job ID can be verified in the IBM Quantum console. Full JSON payloads ship with the licensed QPC buyer package.

Task Tier Layout Backend IBM job ID(s)
T01 — 4-context CO₂ (one submission) T1 4×39Q = 156Q ibm_fez d8atngdmdd1s73b8p4fg
T02 — 12-context finance (one submission) T2 12×13Q = 156Q ibm_fez d8atp7h59p8c73blnl4g
T03 — 6-context cluster T3 2×(3×26Q) ibm_marrakesh d8bjq6j8amns73bi9i00, d8bjq8c7avuc73dpo1f0
T04 — 12-context finance cluster T3 3×(4×13Q) ibm_marrakesh d8bjqek7avuc73dpo1og, d8bjqg2jki0s73aq0nng, d8bjqhs7avuc73dpo1tg
T05 — 24-context logistics T4 6×(4×6Q) ibm_marrakesh six jobs — see buyer JSON

What the runs establish

T01 & T02 — co-resident contexts, one circuit, one measurement

All labelled contexts share one IBM submission and one shot histogram on the full 156-qubit width. That is genuine co-residence in a single run — not twelve separate queue jobs presented as “parallel.” A matched coupled vs decoupled control at identical width and shots is now published for T03-class clusters — see Coupling ablation (bridge ICC + noise mitigation).

T03 & T04 — transjunction gates in-circuit

Cluster jobs include transjunction bridges between contexts inside each submission. The coupling ablation page reports ON vs OFF at 3×26Q on ibm_fez and ibm_marrakesh (1024 shots, bridge-local ICC with QPC readout mitigation).

T05 — hierarchical scale-out

Twenty-four labelled contexts executed as six cluster jobs with classical meta-merge — evidence of orchestration at scale, not a single 156Q flat layout.

Across all tasks

Where the stronger discrimination lives

For reviewers and buyers asking “what proves the architecture, not only the orchestrator?” — see:

Planned extension (not yet on this page): one T03- or T04-class cluster run with transjunctions on vs off, same shots and width — to test whether coupling measurably changes an observable uncoupled blocks cannot reproduce.

Related pages

← Highlights CO₂ multi-context Noise Reducer QQ pilot (K=1 vs K=4)

Not included here: IBM Condor 520Q (account pending); scripts and dry-run assets are in the buyer package under 09_When_Condor_520Q_Access_Exists/.