Two-qubit gate fidelity
99.9%
Median over all 2,016 pairs, interleaved randomised benchmarking.
A fictional brand, designed and built by Get Ranked On Top. Not a real company.
Trapped-ion quantum computing · Boulder, Colorado
Sixty-four ytterbium ions, suspended in vacuum and held perfectly still. Every one connects to every other. The result is a quantum computer whose circuits run as written, and whose answers you can trust.
Apply a gate
State
|ψ⟩ = 1.000|0⟩ + 0.000|1⟩
θ = 0.0° φ = —
Circuit
|0⟩──── no gates applied
¹⁷¹Yb⁺ chain · 16 of 64 ions · spacing ≈ 4.8 µm
Ions held 82 µm above the trap surface at 4 K. Press Entangle to drive a two-qubit gate.
Tessera T64 · system performance
We report medians across the whole machine, never a best-case pair. These are the figures from the benchmark run of 12 October 2026.
Two-qubit gate fidelity
99.9%
Median over all 2,016 pairs, interleaved randomised benchmarking.
Physical qubits
64
Identical ¹⁷¹Yb⁺ ions in a single chain. No fabrication variance, by nature.
Connectivity
All-to-all
Any qubit entangles with any other in one gate. No SWAPs, no routing overhead.
How trapped ions work
Every atom of ytterbium-171 is identical to every other in the universe. That is the quiet advantage behind everything we build.
STEP 01
A single ytterbium atom loses one electron and becomes an ion. Radio-frequency fields on a gold-patterned chip hold sixty-four of them in a line, 82 micrometres above the surface, in a vacuum thinner than interplanetary space.
STEP 02
Each qubit lives in two hyperfine levels of the same ion, split by 12.6 GHz. Tightly focused laser beams rotate any one qubit, and a shared mode of motion lets any two exchange quantum information directly.
STEP 03
A resonant laser makes ions in |1⟩ glow while ions in |0⟩ stay dark. A camera counts photons from all sixty-four at once. Ions are identical by nature, so there is no calibration drift between qubits to chase.
Cloud access
Our Python SDK speaks native circuits and OpenQASM 3. Write the algorithm you mean; the compiler never inserts a SWAP.
from ionfield import Client, Circuit client = Client(project="materials-lab") # token read from IONFIELD_TOKENqpu = client.backend("t64-boulder") ghz = Circuit(4)ghz.h(0)for q in (1, 2, 3): ghz.cx(0, q) # all-to-all: no SWAP chainsghz.measure_all() job = qpu.run(ghz, shots=2000, mitigation="debias")print(job.result().counts)# {'0000': 1003, '1111': 989, '0001': 3, '1110': 5}Compiled circuit · t64-boulder
Roadmap 2026 – 2030
Each milestone is a physics result first and a product second. We publish the result before we ship the system.
2026· now
64 qubits, all-to-all, 99.9% median two-qubit fidelity. Generally available on the cloud since March.
2027
Two trapping zones on one chip, mid-circuit measurement and qubit reuse at full fidelity.
2028
Two traps joined by photonic interconnect. Remote entanglement at 250 pairs per second.
2029
24 error-corrected logical qubits with logical error rates below 10⁻⁶ per cycle.
2030
100 logical qubits across a network of four traps. Chemistry beyond classical reach.
Research
Forty-one peer-reviewed papers since 2019. Every benchmark we quote is reproducible from a public dataset.
IF-TR-2026-014
H. Okafor, L. Brandt, S. Ishikawa, R. Vance
Physical Review Letters · 2026
IF-TR-2026-009
M. Ellison, D. Quintero, K. Halvorsen
Nature Physics · 2026
IF-TR-2025-031
J. Arceneaux, P. Lindqvist, H. Okafor
PRX Quantum · 2025
IF-TR-2025-022
R. Vance, T. Mbeki-Lowe, A. Sørensen-Hart, L. Brandt
Preprint · 2025
Research partners
Halden Institute for Materials
Battery cathode simulation
Arlesford University
Graduate teaching programme
Corvid Therapeutics
Binding-energy estimates
Northwater Grid Cooperative
Unit-commitment optimisation
Kepler Lane Research
Monte Carlo pricing studies
Front Range Quantum Consortium
Open benchmarking
We stopped writing SWAP networks. On Ionfield the circuit we draw on the whiteboard is the circuit that runs, and our error budgets finally close.
What I value most is the honesty of the numbers. The fidelities in the dashboard are the fidelities we measure on Monday morning.
Our graduate students ran their first entangling gate in week two. The documentation reads like it was written by someone who teaches.
Request access
Tell us what you are working on. An applications physicist will reply personally, help you size the problem, and set up your first project.
Questions
Researchers at universities and national laboratories, and engineering teams at companies with a defined workload. We review every request by hand, usually within three working days, and match you with an applications physicist who stays with your project.
Commercial access is billed per system-hour, from $2,400, with reserved blocks available by the quarter. Academic groups receive 40 system-hours per quarter at no cost, and teaching use is always free.
Our Python SDK accepts native Ionfield circuits and OpenQASM 3. The compiler maps any two-qubit interaction directly onto the hardware, so nothing is lost to routing. A noise-faithful emulator runs locally for development.
It is the median fidelity of our entangling gate across all 2,016 possible qubit pairs, measured by interleaved randomised benchmarking and published weekly with full distributions, not best-case pairs.
Circuits and results are processed in our Boulder facility and stored encrypted in the United States. Results are deleted after 90 days unless you choose to retain them. We never train on, inspect or share customer circuits.