Capabilities
What Merlin Quantum does for the buyer, with one proof point each. Hardware means measured on a quantum processor; classical computation means CPUs and GPUs.
We run customer problems at the full size of today's largest quantum processors: up to 256-atom neutral-atom registers and 156-qubit superconducting chips.
Proof pointFull-chip runs on a 156-qubit processor.
Measured on quantum hardwareDevice result · Our own runs on a 156-qubit superconducting processor and on neutral-atom registers of up to 256 atoms.
The same problem runs across superconducting, neutral-atom, trapped-ion and photonic machines, so you are never locked to one vendor.
Proof pointFour hardware types, one method stack.
Measured on quantum hardwareDevice result · Our own runs on superconducting, neutral-atom, trapped-ion and photonic machines.
On a 144-atom quantum-dynamics problem, the processor delivered a scaling law that the quimb classical simulator we ran, with every nearest-neighbour coupling kept, did not converge on at bond dimensions up to 128.
Proof pointMeasured exponent 0.382 ± 0.105 against a predicted 0.386.
Measured on quantum hardwareDevice result · Measured on a 144-atom neutral-atom processor. The classical comparison is our own quimb run, bond dimension up to 128.
A device-quality readout that extracts a processor's own clock to a few parts in a hundred thousand and pinpoints the qubit that is losing information.
Proof pointClock read to 0.0029%.
Measured on quantum hardwareDevice result · Read from the processor's own output with our device-quality readout.
Circuit depth stays constant as the problem grows, so results hold on current hardware without waiting for error correction.
Proof pointThree entangling layers at every width; 126 qubits in a single run.
Measured on quantum hardwareDevice result · A 126-qubit run on a superconducting processor, three entangling layers.
We fix the answer before the machine runs. On an industrial materials-ranking task, the hardware matched the committed ranking exactly.
Proof pointRank correlation 1.0; every energy difference within 0.3%.
Measured on quantum hardwareDevice result · IBM Heron, protocol frozen before the run. Detail in the EUV photoresist case.
AI-guided quantum chemistry on an industrial photoresist cluster at twice the published size, every step within chemical accuracy of its exact reference.
Proof point40 qubits against the published ~20.
Classical computationMeasured on quantum hardwarePublished paper + our computation · The ~20-qubit baseline is the published work on this method; the 40-qubit figure is our GPU computation with hardware sampling.
Proprietary methods shrink the problem before it reaches the machine, so GPU training that took a published 173 hours now takes minutes.
Proof point173 GPU-hours → minutes on one GPU.
Classical computationPublished paper + our computation · 173 GPU-hours is the published training cost of the challenge method; minutes on one GPU is our own computation.
Judged among the best in the world on industrial materials chemistry, in the GIC 2026 Advanced Materials track.
Proof pointAward winner · GIC 2026 · 2nd worldwide
Judged resultChallenge judging · GIC 2026 Advanced Materials track, judged result. GIC challenge page (pqic.org) ↗
How to read these
A capability is only useful if you can trust where it stops.
Decide before you build.
We tell you which option wins, and whether quantum is the reason.
