Reversible elliptic-curve arithmetic
executed gates / 1,464 logical qubits
The resource frontier
The latest circuit sets the public ECDSA.fail Toffoli record at 751,182 executed gates using 1,464 logical qubits. It improves both resources over the previous gate-count record. A separate classical-control optimization produces an accepted product result of 939,169,018 at 1,234 qubits. These are independently scored circuits with different resource objectives.
| Result | Logical qubits | Executed Toffolis | Q × T |
|---|---|---|---|
| Current Toffoli record | 1,464 | 751,182 | 1,099,730,448 |
| Accepted product result | 1,234 | 761,077 | 939,169,018 |
Moving left saves quantum storage; moving down saves gates. The green step connects the 85 nondominated public resource pairs, and the orange point is my current Toffoli record. This is my one remaining point on the current frontier. The minimum-width and minimum-product circuits belong to other participants. The gate axis is logarithmic.
The improvement
The new official Toffoli record is 751,182 gates at 1,464 logical qubits. Compared with the preceding 752,991-gate record at 1,863 qubits, it saves 1,809 gates (0.24024%) and 399 qubits. Its measured product is 1,099,730,448.
The accepted product submission scores 939,169,018 at 1,234 qubits and 761,077 executed Toffolis. Exact classical-control simplifications improve the public product record at upload by 23,446 without adding quantum wires. A later participant has since lowered the global product further.
Both panels start at zero and label the full counts. The reduction in gate count is small, 0.24024%, while workspace falls by 399 logical qubits. These are whole-circuit, officially scored measurements, compared with the immediately preceding gate-count record.
Validation ledger
Both circuits pass their complete unchanged 9,024-shot benchmark packets. Separate fresh-input checks use 2,097,152 shots per source for the Toffoli circuit and 16,777,216 per source for the product circuit. Their declarations, failure counts, nonce searches, and source hashes are retained in the evidence.
The accepted product circuit adds bounded Boolean-identity passes to the inherited public design. They track immutable classical value atoms and simplify controls only when an identity is valid across their full bounded support. Unsupported expressions are discarded conservatively. These classical simplifications add no quantum wires, keeping peak width at 1,234 qubits while reducing executed gate count.
The new Toffoli circuit uses the fixed public 140-row schedule, uniform +2 comparison precision, a plain allocator with clean-pool priming, and generic exact postpasses. It omits the older frozen late cache and source-bound or census-derived cuts from this candidate. Together these changes produce fewer gates with a smaller peak workspace. The public arithmetic and prior postpasses remain attributed; the result is a resource improvement built on that lineage.
The Toffoli circuit's declared census compares 2,097,152 fresh shots per source. Value-failing shots fall from 1,212 to 1,190, phase-failing shots from 816 to 663, and dirty-ancilla shots remain zero. The product circuit has a separate 16,777,216-shot-per-source census. After those checks, each frozen source is replayed on its complete stream-derived benchmark packet and independently scored by the service. Exact transformations preserve the underlying approximate circuit; they do not prove that point addition is correct on every input.
This dated screenshot preserves the earlier product-champion milestone and its crown. It predates the new 20 October results and does not represent the current product standings. The current official receipts and live leaderboard are linked above.
Trace every result