Unitary synthesis challenge (Gold version): is 60% achievable?

Hi,

I’m working on the unitary synthesis challenge, Gold version. My submission is valid and reaches Bronze:

  • saving by size: 0.0%, 8.3%, 9.6%, 9.7%
  • cost: [660, 2904, 12408, 51480]
  • max error: 7.4e-14, no auxiliary wires

Before I continue, I’d like to check something. As I understand the cost model, only RX/RY/RZ carry continuous parameters, and GlobalPhase contributes a single overall phase. A Haar-random 5-qubit unitary has 4⁵ = 1,024 real parameters, so any exact circuit seems to need at least 1,023 rotations. That’s a cost of at least 45,012, or a maximum saving of about 21%. Gold requires at most 22,810.

Could you clarify whether 60% is reachable within the rules, and if so, which part of my reasoning is wrong? For example: are T-gate-based approximate syntheses of the rotations expected, or is there something in the cost model I’ve misread?

Thanks!