Morphogrammatic brickwork and a shallow Bell witness, stepped from IBM Eagle to full IBM Heron r2 width. Public job IDs. Architecture and witness numbers on raw counts.
Same family. Wider chip. Deeper brickwork on the top rung. Bell stays shallow on every rung that carries it.
| 65Q Eagle | 128Q Eagle | 156Q Heron r2 | |
|---|---|---|---|
| Backend | ibm_torino |
ibm_torino |
ibm_kingston |
| Date | 9 February 2026 | 21 February 2026 | 5 September 2026 |
| Brickwork layers | 20 | 20 | 28 |
| Brickwork shots | 2,048 | 2,048 | 4,096 |
| Unique (raw) | 2,048 / 2,048 | 2,048 / 2,048 | 4,096 / 4,096 |
| Compiled size / depth | see 65Q page | 16,617 / 514 | 33,894 / 1,436 |
| Bell pairs F > 0.5 | 59 / 64 | 77 / 78 | |
| Average Bell F (raw) | 0.8148 | 0.8859 | |
| Brickwork job | d64uqfao8gvs73f5246g |
d6d0bffg4t5c7386309g |
dae510dnj4cs73aemir0 |
| Bell job | d6d0gk154hss73b94gc0 |
dae512dnj4cs73aemit0 |
The 65Q brickwork record sits on IBM Quantum summary.
21 February 2026. IBM Eagle ibm_torino · 133 qubits. Same morphogrammatic pattern as the 65Q run, widened to 128.
Job d6d0bffg4t5c7386309g.
64 disjoint neighbor pairs. F > 0.5 means the pair is entangled. Formula: F = (1 + ⟨XX⟩ − ⟨YY⟩ + ⟨ZZ⟩) / 4.
Job d6d0gk154hss73b94gc0.
5 September 2026. IBM Heron r2 ibm_kingston · 156 qubits. Expectation locked before submit. Two separate jobs. Resilience on submit and fetch. Numbers on raw counts.
28 logical layers. 4,096 shots. Compiled circuit larger than the Eagle 128Q rung.
Job dae510dnj4cs73aemir0. Locked bars BL-A1 and BL-A2: job returned, uniqueness recorded on raw counts.
78 disjoint neighbor pairs. Same short H + CX prep as the 128Q witness. Locked bar: at least 55 of 78 pairs with raw F > 0.5.
Job dae512dnj4cs73aemit0. One pair under 0.5: q34–q35, raw F 0.4922. Measured average 0.8859 sits above the Eagle 128Q average 0.8148.
Decision rankings and architecture gaps sit on the Real-World Desks and Highlights.