Blog/Guide

Quantum Device Selection in 2026: A Practical Guide

How to choose the right quantum computer for your workload — with real benchmark data across IBM, IQM, Rigetti, and IonQ.

April 2026 · Based on real hardware experiments

The Problem Every Quantum Team Faces

You have a quantum circuit ready to run. You have access to multiple quantum devices — maybe IBM through Qiskit, IQM and Rigetti through AWS Braket, IonQ through Azure or Braket. The question is simple:

Which device should you run it on?

This is not a theoretical question. In 2026, choosing the wrong device means wasting hours in a queue (IonQ Forte: 5+ hour wait times observed), paying $10+ per run when a $0.65 alternative exists, or getting noise-dominated results when another device would have given usable output. There is no single "best quantum computer." The best device depends on your specific circuit.

What Makes Quantum Devices Different

Every quantum device has a unique profile across three error channels:

Gate Fidelity

How accurately the device executes quantum gates. Errors compound exponentially — a device with 0.5% error per gate behaves very differently on a 3-gate vs. 30-gate circuit.

Coherence

How long qubits retain quantum information. Deep circuits need high coherence; shallow circuits don't care.

Readout Accuracy

How reliably the device measures qubit states. Some devices show 10%+ per-qubit readout error, while others stay below 1%.

Real Data: Same Circuit, Different Devices

We ran three identical circuits on quantum devices from four different vendors. Here are the actual results from April 2026.

4-Qubit GHZ State (Shallow: 3 two-qubit gates)

Device	Vendor	Fidelity	Cost/Run	Queue
IonQ Forte	IonQ	95.3%	~$10.30	5+ hr
IBM Kingston	IBM	96.8%	Free	~2 min
IBM Marrakesh	IBM	94.0%	Free	~2 min
IQM Emerald	IQM	92.8%	$1.75	~9 sec
IQM Garnet	IQM	88.7%	$1.75	~5 sec
IBM Fez	IBM	87.4%	Free	~2 min
Rigetti Cepheus	Rigetti	59.4%	$0.65	~10 sec

Verdict

For shallow circuits, IBM's free tier matches the most expensive hardware. No reason to pay.

4-Qubit VQE Ansatz (Medium: 16 two-qubit gates)

Device	Vendor	Fidelity	Cost/Run	Queue
IonQ Forte	IonQ	56.2%	~$10.30	5+ hr
IQM Emerald	IQM	54.7%	$1.75	~9 sec
IQM Garnet	IQM	53.0%	$1.75	~5 sec
Rigetti Cepheus	Rigetti	39.8%	$0.65	~10 sec
IBM Fez	IBM	28.5%	Free	~2 min
IBM Kingston	IBM	27.8%	Free	~2 min

Verdict

IonQ trapped-ion leads by a hair, but IQM is nearly as good for 6x less money and 2000x faster. IBM's free tier is noise-dominated.

4-Qubit Deep Ladder (Deep: 30 two-qubit gates)

Device	Vendor	Fidelity	Cost/Run	Queue
IBM Kingston	IBM	47.8%	Free	~2 min
IQM Garnet	IQM	37.4%	$1.75	~5 sec
IonQ Forte	IonQ	36.2%	~$10.30	5+ hr
Rigetti Cepheus	Rigetti	31.9%	$0.65	~10 sec
IBM Marrakesh	IBM	29.1%	Free	~2 min
IBM Fez	IBM	27.8%	Free	~2 min
IQM Emerald	IQM	25.6%	$1.75	~9 sec

Verdict

For very deep circuits, IBM Kingston wins on fidelity AND cost. Counter-intuitive but real.

Key Takeaways

There is no single best quantum computer

IBM wins on shallow circuits. IQM wins on medium-depth circuits. The ranking flips depending on your workload.

Free is not always best

IBM's free tier is excellent for shallow circuits, but IQM's $1.75 per run delivers nearly double the fidelity on VQE-style chemistry workloads.

Queue time matters more than you think

IonQ Forte produces the highest shallow-circuit fidelity we measured, but we waited over 26 hours in the normal queue. Hybrid Jobs got us priority access and the run still took 8+ hours. IQM Garnet returns results in 5 seconds.

Readout error is the hidden variable

Rigetti Cepheus showed 10.9% per-qubit readout error — 10x higher than IBM or IQM. Most benchmarks don't report this separately.

You can predict fidelity before running

Using a physics-based model, it's possible to predict whether a circuit will produce usable output on a given device — before spending money or waiting in a queue.

How to Choose: A Decision Framework

Step 1: Know Your Circuit

Count your two-qubit gates. This is the single most important factor.

< 10

Any device works. Use the cheapest.

10 – 20

Device choice matters. Compare vendors.

> 20

Only high-fidelity devices will work.

Step 2: Check Three Numbers

For each candidate device: (1) per-2Q-gate error rate — below 1% is good, (2) coherence depth — must exceed your circuit depth, (3) readout error — below 2% is good, above 5% is a red flag.

Step 3: Estimate Fidelity

Expected fidelity = (1 - gate_error)^gates x coherence_factor x readout_factor

If the result is below 10%, your circuit will produce noise. Pick a different device.

Step 4: Consider Cost and Time

A device that's 20% better in fidelity but costs 10x more and takes 5 hours is not always the right choice — especially during iterative development.

Quantum Device Comparison Table (April 2026)

Device	Vendor	Qubits	Technology	Gate Error	Access	Queue
IBM Fez	IBM	156	Superconducting	~1.0%	Free	2 min
IBM Kingston	IBM	156	Superconducting	~0.7%	Free	2 min
IQM Garnet	IQM	20	Superconducting	~1.7%	$1.75/run	5 sec
IQM Emerald	IQM	5	Superconducting	~1.7%	$1.75/run	9 sec
Rigetti Cepheus	Rigetti	108	Superconducting	~1.0%	$0.65/run	10 sec
IonQ Forte	IonQ	36	Trapped Ion	~0.3%	$10.30/run	5+ hrs
Quantinuum H2-2	Quantinuum	56	Trapped Ion	~0.07%	$125K+/mo	N/A

Gate error rates from Metriq benchmark data. Queue times from direct observation on April 20, 2026.

Tools for Quantum Device Selection

Qlro (Open Source)

Qlro implements the WCPP (Workload-Conditioned Physical Projection) framework — a vendor-neutral scoring system that recommends devices based on your specific workload.

GitHub|pip install qlro|Live Demo

Frequently Asked Questions

Which quantum computer is the best in 2026?

There is no single best quantum computer. The optimal choice depends on your circuit's depth, gate count, and connectivity requirements. For shallow circuits (< 10 two-qubit gates), IBM's free devices are hard to beat. For medium-depth chemistry workloads, IQM devices on AWS Braket offer the best fidelity-per-dollar. For deep circuits requiring very low error rates, Quantinuum H2-2 leads — but at significantly higher cost.

How much does it cost to run a quantum circuit?

It varies by over 1000x across providers. IBM Quantum offers free access to several 156-qubit Heron devices. IQM Garnet on AWS Braket costs approximately $1.75 per 1,000-shot run. Rigetti Cepheus costs $0.65. IonQ Forte costs approximately $10.30. Quantinuum H2-2 requires a monthly subscription starting at $125,000.

How long does a quantum computation take?

The circuit execution itself takes microseconds (superconducting) to milliseconds (trapped-ion). The dominant factor is queue wait time: IBM and IQM return results in seconds to minutes, while IonQ Forte queue times can exceed 5 hours.

Can I predict whether my circuit will work before running it?

Yes. Using a physics-based model that decomposes errors into gate fidelity, coherence, and readout channels, you can estimate the expected output fidelity of any circuit on any device. Tools like Qlro's Circuit Survival Estimator automate this prediction.

What is baseline invariance and why does it matter?

Baseline invariance means that a device's score doesn't change when other devices are added to or removed from the comparison. Many benchmarking platforms normalize against a "baseline" device, causing rankings to shift when the baseline changes. For decision-support tools, this is unacceptable.

Last updated: April 2026. Data from real hardware experiments on IBM Quantum and AWS Braket.

Back to blog