Quantum Hardware Database
IBM Eagle
IBM Eagle was the first quantum processor to exceed 100 qubits (November 2021). Its multi-level wiring let IBM place far more qubits on a chip, and a 2023 experiment on Eagle became one of the most discussed 'utility' results in the field.
Qubit Count
127
Qubit Type
Superconducting (fixed-frequency transmon qubits)
Architecture
Heavy-hex lattice with multi-level wiring (qubits and wiring on separate layers)
Released
2021
Gate Error (2Q)
Roughly 0.7% to 1% median two-qubit error at release; later Eagle revisions improved
Coherence Time
Roughly 100 to 150 microseconds for later revisions (r3 improved coherence over r1)
Operating Temp
About 15 millikelvin
Cloud Access
Yes — IBM Quantum Platform (availability of individual Eagle systems changes over time)
Company
IBM
Overview
IBM Eagle is the processor that took superconducting quantum computing past 100 qubits. Announced in November 2021 with 127 qubits, it was a clear demonstration that IBM's manufacturing and packaging could support larger devices, and it set the template that Osprey, Condor, and Heron built on.
Eagle also has a second claim to fame. In June 2023 IBM and university collaborators published a Nature paper describing a 127-qubit experiment on an Eagle device, with error mitigation, that they described as evidence for the utility of quantum computing before fault tolerance. The experiment is still cited in debates about what noisy machines can do.
History and background
Before Eagle, IBM's biggest processors had 65 qubits (Hummingbird, 2020). The company's roadmap called for roughly tripling qubit counts each year, and Eagle was the first step beyond 100. It entered IBM's cloud fleet in 2021 and 2022, with revisions that improved coherence, such as Eagle r3 in late 2022.
The 2023 utility paper simulated the dynamics of a kicked Ising model, a physics model of interacting spins, on 127 qubits and used zero-noise extrapolation to correct for errors. It prompted a wave of classical simulations from other research groups that reproduced the results using tensor networks and other approximations. The episode is now widely seen as an example of how the line between quantum advantage and classical simulation is contested.
How it works
Eagle uses fixed-frequency transmon qubits and cross-resonance gates, the same basic design as earlier IBM chips. Its innovation lies in the structure of the chip. Qubits sit on one layer, and readout resonators and control wiring are placed on separate layers connected by vertical vias. This three-dimensional arrangement frees space for more qubits on the same chip area.
The heavy-hex layout limits each qubit to two or three neighbors, which reduces the number of ways errors can spread. For algorithms needing other connections, the compiler inserts swap operations. Eagle's two-qubit gate errors at release were around the 0.7% to 1% range, which is high enough that deep circuits quickly lose accuracy without mitigation.
Connectivity
Heavy-hex: each qubit connects to two or three neighbors
Specifications in detail
| Item | Detail |
|---|---|
| Announced | November 2021 |
| Predecessor | Hummingbird (65 qubits) |
| Successor | Osprey (433 qubits), then Heron |
| Notable result | 127-qubit kicked Ising experiment published in Nature, June 2023 |
Key features
- First processor above 100 qubits, at 127
- Multi-level wiring: control and readout lines routed on separate layers beneath the qubits
- Heavy-hex lattice that limits crosstalk
- Platform for the 2023 Nature utility experiment, which used error mitigation on 127 qubits
Limitations
- Higher gate error than later chips such as Heron
- Fixed couplings produce unwanted interactions between neighbors
- Superseded in qubit count by Osprey and Condor, and in quality by Heron
Timeline
| When | What happened |
|---|---|
| 2020 | Hummingbird (65 qubits) sets the previous record for IBM. |
| November 2021 | Eagle becomes the first processor to pass 100 qubits. |
| December 2022 | Eagle r3 improves coherence over the first revision. |
| June 2023 | A 127-qubit Eagle experiment is published in Nature as evidence of utility before fault tolerance. |
What it is used for
- Research on error mitigation techniques at 100-qubit scale
- Spin-model and materials simulations in utility-scale experiments
- Teaching and benchmarking on a widely available superconducting platform
How it compares
- Compared with Heron, Eagle has similar qubit counts but higher error rates and always-on couplings.
- Compared with Google Sycamore (53 qubits), Eagle has more than twice as many qubits but was not designed for the same kind of sampling benchmark.
- Compared with Osprey, Eagle has one third as many qubits and a simpler wiring design.
Technical notes
Eagle's breakthrough was packaging, not physics. Earlier chips routed wiring on the same plane as the qubits, which ran out of room beyond a few dozen qubits. Eagle introduced multiple wiring layers and vertical connections, which let IBM keep adding qubits without crowding the surface.
Why it matters
Eagle changed the scale at which people discuss superconducting quantum computers. It turned 100 qubits from a target into a product, and it showed that the packaging problem could be solved with multi-level wiring.
The 2023 experiment on it also taught the field to be careful with advantage claims. A result that looks beyond classical reach can be matched by a cleverer classical algorithm, so independent verification matters.
Frequently asked questions
What was special about IBM Eagle?
It was the first quantum processor with more than 100 qubits, at 127, and it introduced multi-level wiring that made larger chips practical.
What was the Eagle utility experiment?
In 2023 IBM and partners ran a kicked Ising model on 127 qubits of an Eagle device, with error mitigation, and published the result in Nature. Other groups later showed that classical methods could reproduce the results.
Can I still use Eagle?
Some Eagle-class systems have been available on the IBM Quantum Platform, but the list changes as IBM retires older devices. Check IBM's compute resources for the current options.
How is Eagle different from Heron?
Heron uses tunable couplers, which cut crosstalk and lower error rates by a factor of three to five. Eagle uses fixed couplings and has higher error.
What came after Eagle?
Osprey (433 qubits, 2022), Condor (1,121 qubits, 2023), and Heron (133 then 156 qubits), followed by Nighthawk in 2025.
Keep exploring
- Company profile: IBM
- Related processor: IBM Osprey
- Related processor: IBM Condor
- Related processor: IBM Heron (r1, r2, and r3)
- Related processor: Google Sycamore
This profile is written for educational purposes and summarizes company announcements, published papers, and vendor specification sheets. Last reviewed: September 30, 2026. Figures marked as company-reported have not all been independently verified, and benchmarks differ between vendors, so compare them with care.