“Talking quantum circuits”

Interpretable and scalable quantum natural language processing

September 18, 2024
‍The central question that pre-occupies our team has been:

“How can quantum structures and quantum computers contribute to the effectiveness of AI?”

In previous work we have made notable advances in answering this question, and this article is based on our most recent work in the new papers [arXiv:2406.17583, arXiv:2408.06061], and most notably the experiment in [arXiv:2409.08777].

This article is one of a series that we will be publishing alongside further advances – advances that are accelerated by access to the most powerful quantum computers available.

Large language Models (LLMs) such as ChatGPT are having an impact on society across many walks of life. However, as users have become more familiar with this new technology, they have also become increasingly aware of deep-seated and systemic problems that come with AI systems built around LLM’s.

The primary problem with LLMs is that nobody knows how they work - as inscrutable “black boxes” they aren’t “interpretable”, meaning we can’t reliably or efficiently control or predict their behavior. This is unacceptable in many situations. In addition, Modern LLMs are incredibly expensive to build and run, costing serious – and potentially unsustainable –amounts of power to train and use. This is why more and more organizations, governments, and regulators are insisting on solutions.  

But how can we find these solutions, when we don’t fully understand what we are dealing with now?1

At Quantinuum, we have been working on natural language processing (NLP) using quantum computers for some time now. We are excited to have recently carried out experiments [arXiv: 2409.08777] which demonstrate not only how it is possible to train a model for a quantum computer in a scalable manner, but also how to do this in a way that is interpretable for us. Moreover, we have promising theoretical indications of the usefulness of quantum computers for interpretable NLP [arXiv:2408.06061].

In order to better understand why this could be the case, one needs to understand the ways in which meanings compose together throughout a story or narrative. Our work towards capturing them in a new model of language, which we call DisCoCirc, is reported on extensively in this previous blog post from 2023.

In new work referred to in this article, we embrace “compositional interpretability” as proposed in [arXiv:2406.17583] as a solution to the problems that plague current AI. In brief, compositional interpretability boils down to being able to assign a human friendly meaning, such as natural language, to the components of a model, and then being able to understand how they fit together2.

A problem currently inherent to quantum machine learning is that of being able to train at scale. We avoid this by making use of “compositional generalization”. This means we train small, on classical computers, and then at test time evaluate much larger examples on a quantum computer. There now exist quantum computers which are impossible to simulate classically. To train models for such computers, it seems that compositional generalization currently provides the only credible path.

1. Text as circuits

DisCoCirc is a circuit-based model for natural language that turns arbitrary text into “text circuits” [arXiv:1904.03478, arXiv:2301.10595, arXiv:2311.17892]. When we say that arbitrary text becomes ‘text-circuits’ we are converting the lines of text, which live in one dimension, into text-circuits which live in two-dimensions. These dimensions are the entities of the text versus the events in time.

To see how that works, consider the following story. In the beginning there is Alex and Beau. Alex meets Beau. Later, Chris shows up, and Beau marries Chris. Alex then kicks Beau.

The content of this story can be represented as the following circuit:

Figure 1. A text circuit for a simple story, involving three actors Alex, Beau andChris, who have a number of interactions with one another, making up a story –the circuit is to be read from top to bottom.
2. From text circuits to quantum circuits

Such a text circuit represents how the ‘actors’ in it interact with each other, and how their states evolve by doing so. Initially, we know nothing about Alex and Beau. Once Alex meets Beau, we know something about Alex and Beau’s interaction, then Beau marries Chris, and then Alex kicks Beau, so we know quite a bit more about all three, and in particular, how they relate to each other.

Let’s now take those circuits to be quantum circuits.

In the last section we will elaborate more why this could be a very good choice. For now it’s ok to understand that we simply follow the current paradigm of using vectors for meanings, in exactly the same way that this works in LLMs. Moreover, if we then also want to faithfully represent the compositional structure in language3, we can rely on theorem 5.49 from our book Picturing Quantum Processes, which informally can be stated as follows:

If the manner in which meanings of words (represented by vectors) compose obeys linguistic structure, then those vectors compose in exactly the same way as quantum systems compose.4

In short, a quantum implementation enables us to embrace compositional interpretability, as defined in our recent paper [arXiv:2406.17583].

3. Text circuits on our quantum computer

So, what have we done? And what does it mean?

We implemented a “question-answering” experiment on our Quantinuum quantum computers, for text circuits as described above. We know from our new paper [arXiv:2408.06061] that this is very hard to do on a classical computer due to the fact that as the size of the texts get bigger they very quickly become unrealistic to even try to do this on a classical computer, however powerful it might be. This is worth emphasizing. The experiment we have completed would scale exponentially using classical computers – to the point where the approach becomes intractable.

The experiment consisted of teaching (or training) the quantum computer to answer a question about a story, where both the story and question are presented as text-circuits. To test our model, we created longer stories in the same style as those used in training and questioned these. In our experiment, our stories were about people moving around, and we questioned the quantum computer about who was moving in the same direction at the end of the stories. A harder alternative one could imagine, would be having a murder mystery story and then asking the computer who was the murderer.

And remember - the training in our experiment constitutes the assigning of quantum states and gates to words that occur in the text.

Figure 2. The question-answering task for the language of text circuits as implementable on a quantum computer from [arXiv: 2409.08777]. Above the dotted line is the text we consider. Below are upside-down text circuits which constitute the question we ask. The boxes with words are parameterized as quantum gates. The diagram on the left constitutes one possible answer to the question, and the one on the right the other. Can you figure out what the text is and what the questions are?
4. Compositional generalization

The major reason for our excitement is that the training of our circuits enjoys compositional generalization. That is, we can do the training on small-scale ordinary computers, and do the testing, or asking the important questions, on quantum computers that can operate in ways not possible classically. Figure 4 shows how, despite only being trained on stories with up to 8 actors, the test accuracy remains high, even for much longer stories involving up to 30 actors.

Training large circuits directly in quantum machine learning, leads to difficulties which in many cases undo the potential advantage. Critically - compositional generalization allows us to bypass these issues.

Figure 3. A simplified plot from [arXiv:2409.08777] showing that increasing the sizes of circuits when testing doesn’t affect the accuracy, after training small-scale on ordinary computers. The number of actors correlates with the text size. H1-1 is the name of the Quantinuum quantum computer that was used.
5. Real-world comparison: ChatGPT

We can compare the results of our experiment on a quantum computer, to the success of a classical LLM ChatGPT (GPT-4) when asked the same questions.

What we are considering here is a story about a collection of characters that walk in a number of different directions, and sometimes follow each other. These are just some initial test examples, but it does show that this kind of reasoning is not particularly easy for LLMs.

The input to ChatGPT was:

What we got from ChatGPT:

Can you see where ChatGPT went wrong?

ChatGPT’s score (in terms of accuracy) oscillated around 50% (equivalent to random guessing). Our text circuits consistently outperformed ChatGPT on these tasks. Future work in this area would involve looking at prompt engineering – for example how the phrasing of the instructions can affect the output, and therefore the overall score.

Of course, we note that ChatGPT and other LLM’s will issue new versions that may or may not be marginally better with ‘question-answering’ tasks, and we also note that our own work may become far more effective as quantum computers rapidly become more powerful.

6. What’s next?

We have now turned our attention to work that will show that using vectors to represent meaning and requiring compositional interpretability for natural language takes us mathematically natively into the quantum formalism. This does not mean that there doesn't exist an efficient classical method for solving specific tasks, and it may be hard to prove traditional hardness results whenever there is some machine learning involved. This could be something we might have to come to terms with, just as in classical machine learning.

At Quantinuum we possess the most powerful quantum computers currently available. Our recently published roadmap is going to deliver more computationally powerful quantum computers in the short and medium term, as we extend our lead and push towards universal, fault tolerant quantum computers by the end of the decade. We expect to show even better (and larger scale) results when implementing our work on those machines. In short, we foresee a period of rapid innovation as powerful quantum computers that cannot be classically simulated become more readily available. This will likely be disruptive, as more and more use cases, including ones that we might not be currently thinking about, come into play.

Interestingly and intriguingly, we are also pioneering the use of powerful quantum computers in a hybrid system that has been described as a ‘quantum supercomputer’ where quantum computers, HPC and AI work together in an integrated fashion and look forward to using these systems to advance our work in language processing that can help solve the problem with LLM’s that we highlighted at the start of this article. 

1 And where do we go next, when we don’t even understand what we are dealing with now? On previous occasions in the history of science and technology, when efficient models without a clear interpretation have been developed, such as the Babylonian lunar theory or Ptolemy’s model of epicycles, these initially highly successful technologies vanished, making way for something else.

2 Note that our conception of compositionality is more general than the usual one adopted in linguistics, which is due to Frege. A discussion can be found in [arXiv: 2110.05327].

3 For example, using pregroups here as linguistic structure, which are the cups and caps of PQP.

4 That is, using the tensor product of the corresponding vector spaces.

About Quantinuum

Quantinuum, the world’s largest integrated quantum company, pioneers powerful quantum computers and advanced software solutions. Quantinuum’s technology drives breakthroughs in materials discovery, cybersecurity, and next-gen quantum AI. With over 500 employees, including 370+ scientists and engineers, Quantinuum leads the quantum computing revolution across continents. 

Blog
|
events
November 15, 2026
Super Computing 26
Visit Quantinuum at SC26 – Booth 4838

Quantinuum is heading to SC26 in Chicago, and we'd love to see you at Booth 4838 in the Quantum Junction. Stop by during exhibit hours to catch partner theater sessions throughout the day, see what's new with Helios and Nexus, and talk with our team about how quantum computing is expanding its footprint in the HPC datacenter, including hybrid quantum-classical applications.

In our booth:

Join our experts and partners at our kiosks and for short sessions in our theater running throughout the day. Topics include:

  • Next Gen Enterprise Computing Partners – how leading enterprises are bringing quantum into their computing strategy
  • Error Correction: The Path to Fault Tolerance – where quantum error correction stands today and what comes next
  • Develop for the Fault Tolerant Era with Nexus – building, running, and managing quantum workflows on the Nexus platfor
  • Helios: The Most Accurate Quantum Computer – a closer look at our latest system and what its performance makes possible
  • Startup Partners: Accelerating Quantum Ecosystem Innovation – how startups are building on Quantinuum technology
  • Real Commercial Impact – quantum applications in pharma, chemicals, materials, and finance
Coffee and cookies on us

Need a break from the show floor? Our barista will be serving up coffee, and we'll have fresh cookies every afternoon. Grab a cup, pull up a seat for a theater session, and chat with the team.

Catch Quantinuum in the SC26 program

Our experts are also speaking in the SC26 technical program. Add these sessions to your schedule (all times CST):

Speaking Sessions
November 15th

9:45 – 10:00am | Room: W178a
‍Quantum-HPC Workflows Across Multiple Quantum Computing Platforms: Two Case Studies – part of the 2nd International Workshop for Software Frameworks and Workload Management on Quantum-HPC Ecosystems
Presenting: Philipp Seitz

November 16th

3:30 – 4:00pm | Room: W179a
‍Software for Large Scale Application – part of the 2nd Annual Workshop on Large-Scale Quantum-Classical Computing (LSQCC)
Presenting: Juan Pederson

November 19th

3:30 – 4:00pm | Room: W181abc
‍Bringing Quantum Computing to the HPC Datacenter
Presenting: Kentaro Yamamoto

See you in Chicago

Whether you're exploring quantum for the first time or planning how to integrate it with your HPC environment, our team is ready to talk. Find us at Booth 4838 in the Quantum Junction, November 16–19.

‍

events
All
Blog
|
corporate
November 10, 2026
Reimei Powers Up to Help Japan Lead the Next Era of Discovery  
  • RIKEN and Quantinuum have upgraded Reimei to System Model H2, bringing quantum advantage into sight with their hybrid supercomputing platform.
  • Using the new quantum processor, a Quantinuum-led frontier project team is combining quantum and classical data to map complex material behavior, using Reimei for classically challenging regimes, the JHPC-quantum GPU supercomputer ROQUO and the supercomputer Fugaku for large-scale simulation and AI, and generative models to predict unexplored conditions.
  • A second frontier project is focused on creating a scalable hybrid workflow for biomolecular chemistry, combining quantum computing, supercomputing, and AI to simulate the excited-state properties of complex molecules—with potential applications in drug discovery, advanced materials, and solar energy. Once these projects conclude, Reimei will start regular operations under the Test User Program.

The future of supercomputing just got an upgrade in Japan.

RIKEN and Quantinuum have upgraded the Reimei quantum computer to Quantinuum’s System Model H2, bringing quantum advantage into sight with their hybrid supercomputing platform. Their combination of Reimei, the supercomputer Fugaku, and the ROQUO GPU supercomputer is a world-leading platform that reaches far beyond what any modality can currently offer on its own.

With 56 fully-connected high-fidelity qubits, Reimei is already supporting two frontier projects in quantum materials and biomolecular chemistry. Both are designed to move hybrid computing beyond infrastructure demonstrations and toward high-value scientific applications.

Expanding the landscape of supercomputing

Many scientific problems contain both classically ‘accessible’ and classically ‘inaccessible’ regions.

For example, a classical supercomputer may efficiently explore broad areas of a model’s parameter space but struggle near a phase transition, where correlations grow and the required computational resources increase sharply. In chemistry, classical methods can model much of a molecule accurately but encounter difficulties in small, strongly correlated regions that determine its most important properties. In contrast, quantum computers can often handle these problems deftly, thanks to their native “quantumness”.

We have been working to bring quantum computers together with classical computers, using them as specialized accelerators inside broader heterogeneous computing environments.

RIKEN’s platform is our first instantiation of that vision, bringing together three complementary resources:

  • Fugaku, RIKEN’s classical supercomputer, provides large-scale classical modeling, data processing, and established scientific computing methods.
  • ROQUO, RIKEN’s GPU supercomputer, supports tensor-network simulation, AI training, and massively parallel computation.
  • Reimei, RIKEN’s quantum computer, performs selected calculations involving quantum states and correlations that become prohibitively expensive to represent classically.

Quantinuum Nexus’s hybrid workflow management system coordinates these resources, allowing jobs running on classical infrastructure to submit quantum tasks to Reimei and incorporate the results into a larger calculation.

This foundation has been operational for some time now. A major result from earlier this year used Reimei and Fugaku together to investigate chemical reactions inside biomolecules. Recently, we expanded our work using Fugaku and QPUs: coordinating Reimei and ibm_kobe in a combined VQE–QPE workflow to calculate a simple system’s ground-state energy. In all these cases, quantum hardware augmented the classical supercomputer, with each processor assigned the part of the calculation best suited to its strengths.

The addition of ROQUO now extends this architecture with GPU-accelerated simulation, optimization, and AI. Combined with the upgrade of Reimei to System Model H2, we have launched two frontier projects to push the forefront of scientific computing and prove out Reimei’s capability.

Project 1: GenQAI for materials discovery

The first frontier project brings together quantum simulation, classical tensor networks, and generative AI to investigate many-body physics.

The goal is to discover a sort of ‘treasure map’ for a material: the phase diagram. A phase diagram is a multidimensional map that shows how a material’s properties change when variables like temperature, pressure, or magnetic field are changed. Phase diagrams can hold the secret to exotic properties like superconductivity; can point the way to novel materials for memory devices and energy transfer, or can help to elucidate new medical imaging techniques.

Phase diagrams are extremely valuable, but they can be difficult to calculate or measure. To measure one, you would need a laboratory that could detect material properties at hundreds of points while also generating repeatable parameter changes. Computationally, the problem is similar, but (as mentioned above) one often encounters regions where correlations extend across larger portions of the system, making it incredibly difficult to compute. On top of this, during pre-thermal dynamics or phase transitions, entanglement can grow rapidly, further stretching classical computing capacity.

Reimei is well suited to probing these difficult regimes. We have already used System Model H2 to explore quantum magnetism at scales that frustrate classical approaches. Its high-fidelity operations and flexible qubit connectivity enable the execution of demanding quantum many-body circuits at a scale where matched-accuracy classical simulation can become impractical. But this quantum data is also expensive to generate. The goal is therefore to use it selectively.

ROQUO can perform dense tensor-network simulations across classically tractable regions, parallelizing calculations over model parameters and bond dimensions. This allows Reimei to focus on regimes requiring deep time evolution circuits, where long-range correlations and growing entanglement can sharply increase the cost of classical simulation. Classical data alone is not enough to train a robust generative model across the phase diagram, and that is why quantum data becomes essential.

The team has already collected hardware data for a 6 × 6 two-dimensional spin model, trained a Transformer model for zero-noise extrapolation across the parameter space, and applied active learning to prioritize additional quantum data collection.

Using these results, the team trained a generative AI model capable of learning a parametric representation of the fluctuations and many-body correlations across the combined dataset: the classically reachable and the classically hard regions are both well described by our trained generative model.

Once trained, the generative model is used at inference time on ROQUO to sample the phase diagram at conditions that have not yet been simulated and obtain results for material properties with good accuracy. Step by step, the phase diagram for a material is filled out, generating a ‘treasure map’ of its key behaviors.

For materials researchers, this ‘map’ can add up to a lot. Quantum computers, such as Reimei, could contribute to more accurate predictions of phase diagrams and response functions, helping researchers understand how materials respond to external conditions, a key determinant in electronics, sensing, medical imaging, energy infrastructure, and more.

Project 2: A hybrid blueprint for biomolecular chemistry

The second frontier project applies the same hybrid philosophy to one of quantum computing’s most anticipated applications: the accurate simulation of complex molecules.

Researchers from RIKEN, NVIDIA, and Quantinuum are working together to develop a scalable QPU–GPU–HPC workflow for predicting realistic chemical properties of biomolecular systems. Our initial target is zinc porphyrin, a light-absorbing molecule whose response to light is influenced by its surrounding environment.

While many current projects in quantum computational chemistry are focused on ground state searches, this project focuses on excited states—the electronically energized states involved in processes such as light absorption. These states are typically more complicated to calculate than a molecule’s ground state, but they are essential to understanding how photoactive molecules function in the real world.

As before, the hybrid workflow assigns each part of the calculation to the most appropriate processor:

  • Fugaku and conventional HPC methods represent the chromophore’s larger molecular environment and perform the broader chemistry workflow.
  • ROQUO’s NVIDIA GPUs support optimization, configuration-interaction calculations, AI-assisted circuit design, and spectral reconstruction.
  • Reimei concentrates on strongly correlated electronic regions and searches for important electron configurations that classical methods may miss.

This gives the QPU a focused role: discovering important determinants (candidate electronic configurations) that can enrich the classical calculation. The HPC systems then use those configurations to calculate energies and reconstruct experimentally meaningful properties, including the molecule’s absorption spectrum.

This separation of responsibilities makes the workflow scalable. The QPU is not asked to represent the entire biomolecular environment. Instead, it targets the part of the electronic structure problem where quantum sampling could add information that is difficult to obtain through classical methods alone. To help tackle this, we developed a new hybrid method to solve complicated electronic structure problems.

This approach could provide a blueprint for studying functional biomolecules at the accuracy needed for industrial research. Applications could eventually span drug discovery, biomolecular engineering, advanced materials, and the design of more efficient light-harvesting systems such as solar cells.

A platform for the next era of discovery

Hybrid computing is sometimes described as a future architecture. At RIKEN, these foundations have been operational since 2025 – and they just got a massive upgrade.

Reimei supplies high-fidelity quantum computation. Fugaku provides the scale of a world-leading supercomputer. ROQUO adds accelerated computing and AI. The Nexus workflow management system brings the systems together into coordinated scientific workflows.

Together, they demonstrate a broader principle: the value of quantum computing will not be measured by what a QPU can do in isolation, but by what becomes possible when quantum computers are integrated into the world’s most advanced computing environments.

With Reimei, RIKEN and Quantinuum are building that future now—and building it with Japan.

*Reimei deployment is supported by the New Energy and Industrial Technology Development Organization (NEDO), an organization under the jurisdiction of Japan's Ministry of Economy, Trade and Industry (METI)'s "Development of Integrated Utilization Technology for Quantum and Supercomputers" as part of the "Project for Research and Development of Enhanced Infrastructures for Post 5G Information and Communications Systems."

corporate
All
technical
All
Blog
|
corporate
October 28, 2026
Fidelity First in the Quantum Race

By: Harry Buhrman, Chief Scientist

Quantum computing is often framed as a race to the most qubits, the fastest gates, and the biggest benchmarks.

But a race needs a destination. In the fault-tolerant era, that means solving a valuable problem faster or at a lower cost than the best classical approach—and the time to solution is ultimately what matters most.

In this context, gate speed metrics have received a lot of attention. While gate speed is a meaningful component of the overall quality of a quantum computer, what actually matters most is fidelity. As we all know, speed on its own matters little if we don’t arrive at the correct destination. 

Fidelity has emerged as the most important foundation for success in quantum computing. Higher fidelity means fewer errors in the underlying physical operations. Combined with quantum error correction, higher fidelity can reduce the number of additional qubits and operations needed to make computations reliable. This reduction in overhead can lower costs and shorten the time to deliver a useful answer—even compared with systems that have faster gates but lower fidelity.

Fast Gates Do Not Guarantee Fast Answers

‍Consider a simple example. In an independent-error model, a computation with 10,000 operations, each with an error rate of 10⁻³, or 99.9% fidelity, will experience about ten faults on average that can render the computation useless.

Now make every gate ten times faster without changing its error rate. You still get ten faults. You just get them ten times sooner.

Speed matters, but the relevant measure is the total time required to obtain a trustworthy and useful result. Getting to that point requires performance in many different and connected areas such as state preparation, measurement, error correction, repeated runs, and other work needed to complete the computation.

Gate speed alone cannot establish which system performs best. A fast machine that cannot reliably finish a sufficiently large computation cannot deliver commercial value.

Why Fidelity Changes the Resource Equation

‍Quantum error correction protects information by encoding it across multiple physical qubits. That protection requires resources: additional qubits, measurements, operations, and classical processing. These techniques work better when the underlying physical operations are reliable enough that increasing error protection reduces the logical error rate. 

This requirement is known as the fault-tolerance threshold, the point below which physical error rates are low enough that adding protection reduces logical errors. Faster gates can shorten the computation, but they cannot replace the physical fidelity needed to cross that threshold. 

At Quantinuum, our latest error-correction results on Helios demonstrate this principle. Using our error-correction architecture, logical memory errors fell exponentially as we increased protection. Across the tested code sizes, the fitted error reduction was about twelvefold for each increase of two in code distance. In the largest quantum-memory experiment, the estimated failure rate was about five in a million per error-correction cycle.1 This demonstrates how high-fidelity physical operations and effective error correction work together to make quantum information more reliable as protection grows.

The next question is what that reliability enables. How large a computation can a system complete, and how quickly can it deliver the result?

A New Performance Benchmark

The Quantum Universal Operations Performance System, or QUOPS, provides a standardized way to assess both. This architecture-neutral benchmark gauges performance by measuring computational “reach” and speed.2

“Reach” measures how large a benchmark circuit a system can execute successfully. Speed measures how quickly it executes that circuit, in effective quantum operations per second. QUOPS uses wall-clock time and accounts for extra samples and discarded runs caused by noise.

Together, these measures provide a common basis for comparing what different quantum systems deliver and tracking progress toward better time to solution.

Quantinuum’s H2 and Helios systems executed substantially larger benchmark circuits than Google’s Willow and IBM’s Boston. These superconducting systems sometimes achieved higher effective operation rates, but on smaller circuits, which means that they are not computationally comparable.

Superconducting systems can use error mitigation to tackle larger computations, but at a steep cost in speed. The QUOPS study estimates that the additional repetitions can make effective processing thousands of times slower. Some computations remain beyond the system’s reliable reach, even with mitigation. In those cases, faster gates may deliver no useful answer at all within practical limits on time and cost.

Useful applications require speed on much larger circuits. For example, the study estimates that factoring RSA-2048 or calculating FeMoco’s ground-state energy within five days would require roughly 100,000 times the computational reach demonstrated by today’s physical-qubit processors. Some machines approach the estimated speed requirement, but only on small circuits. The challenge is to sustain that speed at the required scale, including the additional work needed for error correction.

Larger uncorrected circuits accumulate more errors. Protecting them requires extra qubits and time. Reliability and speed must therefore scale together, and QUOPS offers a way to track both. 

Higher Fidelity Makes Larger, Reliable Computations Possible

Improving fidelity can reduce the extra qubits and operations needed for error correction, lowering the time, cost, and resources required for useful computation. The benefit comes from reaching the required reliability with less protection, leaving more resources available for the computation itself.

Helios brings together the physical performance and error-correction capabilities needed to pursue those savings. In our paper published in Nature we showed that the system achieved an average single-qubit gate fidelity of 99.9975% and average two-qubit gate fidelity of 99.921% across its operational zones, while scaling the trapped-ion architecture to 98 fully connected physical qubits.3

The latest color-code experiments build on that physical performance. Alongside the logical memory results, protected two-qubit Clifford operations also became more reliable as protection increased. This connects improvements in the underlying hardware to improvements in logical computation.

A flexible error-correction architecture creates further opportunities to improve efficiency. Our recent Helix and color-code demonstrations used the same Helios hardware to explore different balances of accuracy and physical-qubit requirements. That flexibility allows us to evaluate which codes best suit different computational tasks as systems scale.

However, we continue to push ahead and accelerate the development of machines, such as Apollo, where quantum utility comes into play at scale. Decoding in the new color-code benchmarks was performed offline, and we are now turning our attention to integrating error correction, magic-state resources and real-time control into longer, reliable computations.

While that work continues, today’s systems can already contribute to scientific research. NISQ systems have advanced algorithms and hardware, enabled certified randomness, and supported studies of quantum magnetism and superconducting pairing correlations. These results leave room for useful applications before fault tolerance, even as error correction provides the scalable path to larger, reliable computations.4

The goal is to combine these capabilities so that larger computations can run reliably within practical limits on time and resources. Fidelity provides the foundation, and optimizing the full system turns that foundation into better time to solution.

Fidelity First. Then Optimize Time to Solution.

‍Scaling quantum computation requires continued progress in fidelity. As physical operations move further below threshold, increasing error protection can suppress logical errors more effectively, enabling longer, more demanding computations.

Speed must improve alongside that reliability. Advances in transport, cooling, measurement, compilation, and parallel execution can reduce the time and total cost of solving useful problems. These improvements become increasingly valuable as systems gain the reliability needed to complete larger computations.

Our latest Helios results are proof points on the path to Apollo, our planned universal fault-tolerant quantum computer, expected by 2029. The goal is to translate these gains into systems that can solve industrially relevant problems faster or at lower cost than the best classical approach.

Industrial partnerships help connect that progress to problems that matter, from understanding high-temperature superconductivity to how molecules interact to accelerate drug discovery. They bring practical research questions into the development process and help identify where improving quantum capabilities could deliver useful results beyond the reach of classical computing alone.

Our motto is simple: fidelity first, with speed built on that foundation.

References & For Further Reading

‍[1] A. Lavasani et al., "Quantum error correction well below the color code threshold," arXiv:2610.24077 [quant-ph] (2026). doi:10.48550/arXiv.2610.24077.

[2] T. Proctor et al., "Benchmarking the computational power of quantum computers," arXiv:2609.12146 [quant-ph] (2026). doi:10.48550/arXiv.2609.12146.

[3] A. Ransford et al., "A 98-qubit trapped-ion quantum computer with all-to-all connectivity," Nature 655, 81-86 (2026). doi:10.1038/s41586-026-10676-4.

[4] M. Liu, R. Shaydulin, P. Niroula et al., "Certified randomness using a trapped-ion quantum processor," Nature 640, 343-348 (2025). doi:10.1038/s41586-025-08737-1.

     S. Dasu et al., "Computing with many encoded logical qubits beyond break-even," arXiv:2602.22211 [quant-ph] (2026). doi:10.48550/arXiv.2602.22211.

      E. Granet et al., "Superconducting pairing correlations on a trapped-ion quantum computer," arXiv:2511.02125 [quant-ph] (2025; revised 2026). doi:10.48550/arXiv.2511.02125.

corporate
All