Start with a simulator to debug and explore a circuit; choose a QPU only when you need to test hardware execution or behavior. For ideal simulation, AWS’s starting point is a local simulator below 18 qubits, workload-dependent choice at 18–24 qubits, and an on-demand simulator above 24. For noise simulation, the corresponding guidance is local below 9 qubits, workload-dependent at 9–12, and DM1 above 12. These are AWS rules of thumb—not runtime guarantees. The right choice also depends on circuit operations, host capacity, device compatibility, queues, and cost.
Choose by what you need to learn
Use the option that matches the experiment, rather than selecting by qubit count alone:
As an Amazon Associate I earn from qualifying purchases.
- Debugging or exploring an ideal circuit: Begin with a local state-vector simulator if your computer has sufficient memory and compute.
- Simulating noise: Use a density-matrix simulator such as local
braket_dmfor small circuits or managed DM1 for larger ones within its limits. - Scaling simulation or running many workloads: Consider managed on-demand simulation, bearing in mind its per-task latency.
- Testing actual hardware behavior: Submit to a suitable QPU after checking its paradigm, gate set, connectivity, availability, queue, limits, and current price.
- Modeling an analog Hamiltonian system: Consider QuEra’s Analog Hamiltonian Simulator only if the problem fits its Hamiltonian, register, and control-field formulation; it is not a general-purpose gate-based circuit device.
A simulator can reveal programming or configuration errors without QPU execution charges, but its output is not guaranteed to match hardware results. AWS describes the distinctions between simulators and devices in How Amazon Braket works.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Use AWS’s qubit ranges as a first filter
AWS’s simulator comparison offers the following starting guidance. Treat it as a screening aid: actual fit depends on circuit shape, operations, host resources, workload volume, and cloud task latency.
#1 Best Overall
| Simulation goal | AWS starting guidance | How to apply it |
|---|---|---|
| Ideal or standard circuit simulation | Fewer than 18 qubits: local simulator. 18–24 qubits: choose based on workload. More than 24 qubits: on-demand simulator. | Check available host memory and compute for local runs. For many small cloud tasks, account for per-task latency. |
| Noisy circuit simulation | Fewer than 9 qubits: local simulator. 9–12 qubits: choose based on workload. More than 12 qubits: DM1. | Noise simulation has higher resource demands; verify the specific simulator’s limit and the circuit’s noise model. |
AWS’s published recommendations are in Compare Amazon Braket simulators. The ranges do not promise a particular runtime or prove that a circuit will fit. AWS says SV1 runtime increases linearly with gate count and exponentially with qubit count; DM1 generally scales linearly with operations and exponentially with qubits. Benchmark the circuit you actually intend to run.
Choose the right simulator
Local state vector: braket_sv
The local SDK state-vector simulator is useful for rapid prototyping and small ideal circuits. AWS documents it for circuits up to 25 qubits, depending on the host machine’s hardware. Local execution avoids managed-task overhead, but you are limited by the memory and compute available on that computer.
Local density matrix: braket_dm
For small noisy circuits, the local density-matrix simulator runs in the SDK environment. AWS documents it for up to 12 qubits, depending on host hardware. Noise modeling can require substantially more resources than ideal state-vector simulation, so the nominal qubit limit is not a guarantee that a particular circuit will run comfortably.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallRank #2
Managed state vector: SV1
SV1 is an on-demand state-vector simulator for ideal simulation. AWS’s getting-started guide describes simulations up to 34 qubits. It is always available and can process multiple circuits in parallel. AWS notes that shots have a relatively small effect on runtime compared with qubit and operation counts. The larger documented capacity can help when local resources are insufficient, but cloud submission adds task latency.
Managed density matrix: DM1
DM1 is AWS’s managed density-matrix simulator for noisy circuits, documented up to 17 qubits. AWS’s simulator-submission guide lists a six-hour maximum runtime, a default of 35 concurrent tasks, and a maximum of 50 concurrent tasks. Check the current AWS limits before planning a large batch; service limits can change, and an individual circuit can hit resource or runtime constraints below the stated qubit ceiling.
Tensor-network simulation: TN1
The available AWS comparison details do not establish enough current information to recommend TN1 for this choice. If a tensor-network approach appears relevant to your circuit, consult the current supported-device documentation and verify its limits and suitability before committing a workload.
AWS’s overview of simulator submissions and limits is available in Submitting quantum tasks to simulators.
Check whether your circuit fits a QPU
AWS documents gate-based QPUs from AQT, IonQ, IQM, and Rigetti, as well as QuEra’s Analog Hamiltonian Simulator. Provider and device inventory can change, so use the live Amazon Braket device documentation rather than assuming a particular device is currently available.
For a gate-based device, inspect its properties before submission:
- Paradigm: Confirm the device runs the kind of computation your circuit represents. An analog Hamiltonian workflow is not interchangeable with a gate-based circuit.
- Supported gates: These are operations the QPU accepts. A gate your circuit uses may need decomposition into operations the device can execute.
- Native gates: This subset can map directly to the device’s control pulses. A compiler may translate other supported gates into native ones.
- Connectivity: Compare the device’s qubit connection graph with the interactions in your circuit; mapping can require routing or additional operations.
- Shots and task limits: Confirm that your requested repetitions and task fit the device’s current constraints.
- Availability and queues: Check the device’s status, execution windows, and queue depths before planning a run.
The local simulator accepts a broader gate set and some OpenQASM features that may not be accepted by a QPU or another simulator. A successful simulator run therefore does not establish that the same circuit can be submitted unchanged to hardware. AWS’s QPU example explains compilation to device-native gates and mapping abstract circuit qubits to physical qubits: Example: Submitting a quantum task to a QPU.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Plan around QPU availability and queues
QPU execution windows and status vary by device. QPU and on-demand simulator tasks are queued; QPUs have limited capacity, and actual execution time depends on other workloads and device availability. In the console, check each device’s status, availability windows, and quantum-task and hybrid-job queue depths. The SDK also exposes queue depth and task queue position.
A device shown as offline may be in maintenance, undergoing an upgrade, or recovering operationally. Queue depth is useful for planning, but it is not a guaranteed completion-time estimate. AWS says QPU tasks can be submitted at any time, including outside an execution window, and will wait for the device. The SDK’s documented five-day default polling timeout is a client-side wait setting, not a promise that a queued task will complete within five days. See When will my quantum task run?.
Best Value
Estimate cost before committing to hardware
Amazon Braket has no upfront commitment, but usage is billed. Simulator and QPU charges depend on the service and workload, so check the current Braket cost tracking and pricing guidance rather than relying on an old example. The SDK Tracker and console estimates can help assess a workload, but estimates may differ from the bill and may not include other AWS service costs or discounts.
AWS offers optional per-device spending limits for QPU tasks. Its documentation says those controls do not cover simulators, managed notebooks, Hybrid Job EC2 costs, or Braket Direct reservations. AWS also recommends billing alerts through AWS Budgets. Account for those exclusions and any associated AWS services when deciding whether a hardware run is worth the time and expense.
Use shots and a simulator-to-QPU workflow deliberately
A shot is one repeated execution and measurement. More shots generally improve statistical precision, but require more repetitions and can affect cost and runtime. Choose a count suited to the measurement accuracy your experiment needs rather than increasing it by default.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors- Run locally when practical: Use a local simulator to check circuit construction and basic behavior while the workload fits your host.
- Move to managed simulation when justified: For ideal simulation, consider SV1 as the workload exceeds local capacity; for noise modeling, consider DM1 within its documented limits. Account for cloud task latency and concurrency.
- Validate hardware compatibility: Check the target QPU’s supported and native gates, connectivity, paradigm, shot and task limits, current status, availability, and queue.
- Estimate the full run: Set a shot count based on needed precision, review current device pricing and task estimates, and account for related AWS costs.
- Submit the hardware experiment: Reserve QPU execution for questions that require hardware results or device-specific behavior; simulator output is not a substitute for those measurements.
For terminology such as tasks, shots, and devices, see Amazon Braket terms and concepts.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
