Atlas · Cost model · TRIAGE-COST

What a QPU job would cost — audited.

Every other tool answers "run my method." Atlas also answers the money question: what would this circuit cost on a real QPU, and what does the classical route cost instead? The number is built from a formula you can see, with assumptions you can edit — labeled published / measured / derived, never a black box.

TRIAGE-COST v3 · pricing snapshot ~2026 (verify at the vendor links below) · deploy/app/economics.py
The defensible figure. For a tractable circuit a laptop settles in milliseconds, the avoided QPU job at casual precision (ε=1e-2, ~10,060 shots) costs ≈ $805 on IonQ Forte — derived exactly: $0.30 task + 10,060 shots × $0.08/shot, published Braket pricing. At research-grade precision (ε=3e-3, ~112k shots) the same job is ≈ $8,900. Open the panel and recompute — that is the point.

1 Shots from precision

To pin an expectation value to additive error ε you need shots that beat shot-noise (error ~ 1/√N):

N_base = ⌈ 1 / ε² ⌉

ε=1e-1 → 100 · ε=1e-2 → 10,000 · ε=3e-3 → ~111,111. derived (textbook shot-noise). Vendor floors apply (e.g. IonQ error-mitigation minimum 2,500 shots).

2 Mitigation overhead — the sensitive heart

Real hardware is noisy, so a usable expectation value needs error mitigation, whose sample overhead grows exponentially with circuit volume × noise (probabilistic error cancellation):

N = N_base · γ², γ² = e^(4·λ), λ = ε_g · n₂q

published bound — Quek/Eisert arXiv:2210.11505, Takagi arXiv:2109.04457. n₂q = two-qubit gate count, ε_g = per-gate error (next section).

Where a reviewer will push — and we agree. This exponent is the single most assumption-sensitive number in the whole panel: small changes in ε_g or the mitigation model move the QPU cost by orders of magnitude. That is exactly why Atlas exposes ε_g and the formula as editable and sourced — the honest answer to "why this model?" is "PEC, published, and you can swap it," not "trust us." We label it published, never measured.

3 Per-device gate error ε_g — the calibration

The exponent rides on ε_g, so Atlas uses per-device values, not one generic number:

Deviceε_g (2-qubit)Provenance
IBM (Heron r2)2.0e-3measured — Atlas's own ibm_kingston RB (2.02e-3 isolated; 3.42e-3 EPLG layered)
Rigetti5.0e-3published — 99.5% 2q fidelity
IQM (Garnet/Emerald)5.0e-3published — 99.51%
IonQ (Forte/Aria)3e-4vendor-claimed — no hard public 2q figure; optimistic, flagged uncertain

The IBM number is measured on our own hardware runs — that calibration data is part of the moat, like the corpus. The rest are published/claimed and editable.

4 Two pricing families — billed how the vendor bills

Most naïve estimators invent a $/shot for IBM. IBM doesn't bill per shot — it bills per minute of QPU time. Atlas branches:

per-shot (AWS Braket): cost = task_fee + N · per_shot per-minute (IBM): cost = runtime_min · rate_per_min

published tariffs — AWS Braket (IonQ $0.08/shot + $0.30/task; Rigetti $0.000425; IQM/AQT/QuEra) · IBM plans (Open free · Pay-As-You-Go $96/min · Flex $72/min · Premium $48/min). Verify: AWS Braket pricing · IBM Quantum plans.

Two honest caveats. (1) Pricing is the most perishable data on the panel — published ~2026, snapshot-dated, verify at the links. (2) IBM "Open" is genuinely free ($0/min, quota-limited); plan minimums (Flex's 400 min/month) are monthly commitments and are not applied per-circuit (an earlier build did, inflating Flex — fixed).

5 The classical side — measured, not estimated

The whole verdict is a comparison, so the classical cost is measured, not guessed: Atlas times the real MPS/tensor contraction (quimb) for the circuit and prices the wall-clock at a cloud-CPU rate. The reference machine is a dev Apple M4 (24 GB) — see the per-estimator ceilings on the Evidence page.

avoided $ = QPU_cost − classical_cost (when a classical route is certified)

A laptop that settles the circuit in milliseconds at near-zero cost, versus the QPU job above, is the "don't spend" verdict in dollars.

6 Worked example (reproducible)

IonQ Forte, a small tractable circuit a laptop runs in ms:

Precision εShots (incl. mitigation)QPU $/job
1e-1 / 3e-22,500 (vendor floor)$200.30
1e-2 (casual)~10,060≈ $805
3e-3 (research-grade)~111,781≈ $8,900

Every row is recomputable from §1–4 with the published pricing; the panel lets you change vendor, ε and campaign size and watch the number move.


Where the moat actually is (stated plainly). Not the formulas — shot-noise, the PEC overhead and the pricing tables are all public, and anyone could code an estimator. The defensible edge is the same one as the rest of Atlas: (1) the integration — measured classical cost vs calibrated QPU cost as a single spend / don't-spend verdict (vendor calculators only price their own hardware; none tell you a laptop does it for free); (2) measured device calibration (ε_g from our own QPU runs); (3) per-shot-vs-per-minute modeled correctly; (4) full auditability — every term labeled and sourced. The page you are reading is the moat: a cost number you can recompute, versus a black-box calculator you can't.
Open the cost panel → How the verdict is produced