1
quantum and quantum-inspired software; advanced computation
No published assessmentDirectory inclusion is separate from a score.
The quantum company directory.
QRank
Search by name or technology. Open a profile for its sources and, where available, its QRank score.
Editorial research assessment, not a performance benchmark or stock rating.
392 listed · 192 research profiles · 77 currently assessedWhat these counts mean
Whole-directory coverage across 8 sectors. The results below follow your selected sector and filters.
- Listed
- Included in the directory. A listing is not a score or endorsement.
- Research profile
- A dedicated page with sources and context. It may still be unscored.
- Currently assessed
- A complete published assessment eligible for the current ranking under Method 1.0. Reviews older than 90 days are excluded.
These groups overlap: an assessed organization is also listed. Score edition .
About QRank
Find a company. Follow the evidence.
Search quantum organizations by name, technology or application. Find official websites and, where reviewed, sources, access routes and limitations. Listed and assessed profiles are clearly distinguished.
Software & applications
Which tools turn quantum research into something people can use?
22 assessed in software & applications · Score edition 23 Sept 2026 · Method 1.0
Your question: Which organizations match software & applications and my filters?
Search companies, technologies or applications. Arrow keys explore suggestions. Escape keeps your search and closes suggestions.
Directory · 73
Full company research →
A
Industrial-materials company researching optical and electronic materials in the joint program.
No published assessmentDirectory inclusion is separate from a score.
A
Automotive and energy-products company researching superconducting materials in the six-company program.
No published assessmentDirectory inclusion is separate from a score.
A
quantum algorithms for life sciences; noise handling
No published assessmentDirectory inclusion is separate from a score.
A
85.0 /100 · editorial assessment
Provides a common access layer for supported processors, classical simulators and jobs combining quantum tasks with classical computing.
Limitation: A shared cloud interface does not imply shared gate sets or equal performance. Analog Aquila programs differ from gate-model circuits; the cited experiments establish use, not commercial speedup.
85.0/ 100
External corroboration
External sources support at least half of the weighted assessment. They can include collaborators; this does not imply independent replication or technical superiority.
How is this score calculated? for Amazon Braket
This score reflects documented maturity using the software & applications weighting.
- Capability 30% weight
- 3 / 4 · 22.5 points
- Delivery 25% weight
- 4 / 4 · 25 points
- Use in practice 30% weight
- 3 / 4 · 22.5 points
- Integration 15% weight
- 4 / 4 · 15 points
Strength: Current device discovery, calibration queries and managed hybrid-job resources.
Limitation: Access does not make different processors interchangeable. Queues, calibration, gate support, shots and classical execution affect results and total cost.
External corroboration · Assessment reviewed
Score edition · Method 1.0. Editorial opinion within this sector, not a probability or performance benchmark.
A
application marketplace; orchestration; PlanQK platform; PlanQK is now part of Kipu Quantum Hub
No published assessmentDirectory inclusion is separate from a score.
A
quantum application development; optimization platform
No published assessmentDirectory inclusion is separate from a score.
A
Platform for quantum datasets, reproducible benchmarks and application challenges; quantum hackathons and tools connecting researchers with enterprise use cases.
No published assessmentDirectory inclusion is separate from a score.
A
Industrial-materials company researching battery-related materials with computational and experimental feedback.
No published assessmentDirectory inclusion is separate from a score.
A
85.0 /100 · editorial assessment
Helps developers write quantum programs, submit supported workloads to cloud providers and explore the resources future error-corrected machines would need.
Limitation: That study models error-corrected resources; it does not execute a commercially useful fault-tolerant application. Estimator outputs change with the chosen hardware and error assumptions.
85.0/ 100
External corroboration
External sources support at least half of the weighted assessment. They can include collaborators; this does not imply independent replication or technical superiority.
How is this score calculated? for Azure Quantum
This score reflects documented maturity using the software & applications weighting.
- Capability 30% weight
- 3 / 4 · 22.5 points
- Delivery 25% weight
- 4 / 4 · 25 points
- Use in practice 30% weight
- 3 / 4 · 22.5 points
- Integration 15% weight
- 4 / 4 · 15 points
Strength: A documented development environment connects quantum programming, hardware submission and resource modeling.
Limitation: Resource estimates describe assumed future architectures, not delivered fault-tolerant execution. Cloud performance remains provider- and workload-specific.
External corroboration · Assessment reviewed
Score edition · Method 1.0. Editorial opinion within this sector, not a probability or performance benchmark.
B
Computational chemistry and molecular-modeling software; quantum algorithms and error-correction research; current drug-discovery workflows also run on classical hardware.
No published assessmentDirectory inclusion is separate from a score.
b
blueqat develops tools spanning quantum circuit simulation, cloud interfaces and lower-level device control. Its current platform connects developer SDKs and APIs with tensor-network methods and quantum-circuit sampling, while the company also pursues semiconductor quantum-hardware research. These tools support experiments at the intersection of quantum computing, optimization and machine learning, with access depending on the chosen service. Based in Tokyo, Japan.
Limitation: Tensor-network calculations run classically. API access, simulation capability and a proprietary hardware roadmap should not be treated as the same demonstrated quantum capability.
No published assessmentDirectory inclusion is separate from a score.
Since your review
Scope: Software & applications. Reading a page never marks its records reviewed.
Keep a baseline and explicitly mark the evidence you have reviewed. This optional journal stays in this browser.
Coverage, methodology & sector field guide
22 assessed / 73 listed
Directory inclusion is separate from assessment.
A field guide to this sectorKnow what to compare
Start with a workload and a baseline
Circuit design, execution tools and application research play different roles. A result on a classical simulator is distinct from an experiment on quantum hardware.
- Compilation
- Translates a program into instructions for a target device.
- Hybrid workflow
- Combines classical and quantum processing.
Three questions worth asking
- Does the tool support your intended hardware and workflow?
- Does the result count sampling, mitigation and classical processing costs?
- Can someone outside the company reproduce or use the workflow?
Four checks before comparing a hardware claim
Use these questions when reading a company’s sources. They explain comparisons; they do not assign or change a QRank score. Every numerical example below is synthetic.
Best device or typical device?
What population does the headline number describe?
A selected result and a median answer different questions. Compare the same statistic, selection rule, measurement method and population before ordering two devices.
Synthetic sample: five coherence times are 40, 50, 60, 70 and 100 microseconds. The best is 100; the median is 60. Both describe this sample, but neither establishes the performance of every qubit or a useful circuit.
Ask for: Find the sample size, tested device set, excluded measurements, distribution and calibration date. If only the best value is disclosed, leave the median unknown.
NIST: defining mean, median and other measures of location
Link to this checkTested alone or operating together?
Were nearby qubits active during the measurement?
Operating a gate alone and operating it alongside other gates can expose different unwanted interactions. Simultaneous benchmarking explicitly tests an operating context that an isolated measurement does not establish.
Synthetic comparison: an error estimate rises from 0.2% in isolation to 0.8% with neighboring operations active. That is a 0.6 percentage-point change, or four times the original estimate. Without the method and uncertainty, those two numbers do not diagnose the cause.
Ask for: Find the active qubit set, simultaneous operations, benchmark protocol, uncertainty and calibration window. Do not compare an isolated best case with a simultaneous average as though they were the same test.
Gambetta et al.: individual and simultaneous randomized benchmarking
Link to this checkError per gate, per cycle or per task?
What exactly counts as one opportunity for error?
A physical-gate error, a logical-cycle error and an application failure rate use different units and protocols. Keep the operation, code and decoding conditions with the result. A smaller percentage with a different denominator does not settle a comparison.
Synthetic classical event model: if ten independent operations each have a 1% chance of a flagged failure, the chance of at least one flag is 1 − 0.99¹⁰, about 9.56%. This is not a conversion from quantum gate infidelity to circuit failure: coherent errors, correlations and cancellation can break that interpretation.
Ask for: Find the error definition, benchmark, number of repetitions and confidence interval. For logical results, also find code distance, cycle definition, decoding method and any discarded trials.
Magesan et al.: the assumptions of interleaved gate benchmarking
Link to this checkFast operation or fast complete job?
Where does the reported stopwatch start and stop?
Circuit execution is one part of a complete experiment. Preparing the job, waiting for access, repeating measurements and processing the output can each take time. Device usage and elapsed time are also different accounting quantities.
Synthetic job: 1,000 repetitions of a 50-microsecond gate sequence contribute 0.05 seconds. Add 0.25 seconds for readout/reset, 2 seconds of preparation, 10 seconds waiting and 1 second processing: the complete toy job takes 13.3 seconds.
Ask for: Find included stages, repetitions, error-mitigation work, queue treatment and output-quality requirement. Compare equivalent completed tasks before interpreting a speedup.
IBM Quantum: workload usage and execution-time components
Change the assumptions in the experiment-time lesson →
Link to this checkBackground sources checked 2026-10-16. The examples are QubitWire teaching calculations, not results from the cited experiments.
Common QRank questions
- Does the highest score mean the fastest computer?
- No. QRank assesses documented maturity within a sector. Performance depends on the product, workload and test conditions.
- Why does a listed organization have no score?
- A listing and a complete assessment are different. No published assessment means no score has been published; it does not mean poor performance. An expired assessment can remain in a profile as history.
- Does a larger qubit count mean better performance?
- No. Physical and error-corrected logical qubits describe different things. Operations, errors and the resources needed for the task matter too.
- Do more sources mean stronger proof?
- Several pages can describe the same experiment. Company sources can document useful results; external authors may be collaborators. Neither link counts nor external-source badges establish independent replication. The evidence list counts distinct assessment and profile-research URLs; guides can link to additional context.
- Does a new source update the score?
- No. Research context and assessment reviews are separate. Scores change only when the criterion inputs are reviewed and a change is recorded. Source checks and formatting edits do not renew an assessment date.
- Why is an assessment date older than the research?
- A profile can include later context while retaining its earlier score review. A review is due after 30 days; assessments older than 90 days leave the current ranking. The date does not promise a future review.
- How do I check access and pricing?
- Use the linked official route for the named product. Plans, regions, eligibility and features can change; a simulation, pilot or partner integration does not establish public hardware access.
- How can I suggest a correction?
- Use a claim’s correction link and include the product, source location and proposed change, or email contact@qubitwire.com. Submitting evidence does not automatically change a score.
What matters in this sectorView scoring weights
Distinguishes simulations, hardware experiments and production use. Quantum advantage is never inferred from a product launch.
30%Capability
Application results with resources, baselines and limitations stated.
25%Delivery
An accessible, maintained product or service with documentation.
30%Use in practice
Specific external trials or repeated use with reported results.
15%Integration
Hardware compatibility, reusable tools and developer support.
Browse all 192 research profiles by sector
Read, question, follow the source
What would the next piece of evidence show?
Company guides trace the path from an announcement to a measured component, accepted system or exact product certification.
Company evidence guideIQM: when is an installed system accepted?Follow an installed quantum system from supplier delivery to customer acceptance.
It identifies the traded instrument and its relationship to the underlying shares. Readers should use that instrument when checking issuer filings or a licensed quote provider; a warrant or Helsinki share line needs its own identity and market metadata.
Full company guide, sources and limits →Company evidence guideXanadu: what does a foundry partnership establish?Connect the new foundry agreement to the measurements that would establish manufacturing readiness.
It establishes the agreement described by the parties. Production readiness needs additional evidence about fabricated devices, yield, repeatability, packaging and accepted system performance. The record deliberately treats those as separate steps.
Full company guide, sources and limits →Company evidence guidePasqal: what runs on an analog quantum processor?The cited issuer identity and the task a reader can execute using an analog platform.
Yes, as transaction history with its original date and status. Use the later completed-transaction and reporting evidence to distinguish the completed state from the earlier proposal.
Full company guide, sources and limits →Company evidence guideAlice & Bob: what can researchers measure on the testbed?Use the Helium announcement to ask precise questions about encoded experiments and partner access.
No. The complete encoded operation still has to account for the error processes that remain, including those associated with control, measurement and correction. The value of a biased-noise architecture depends on how those parts work together.
Full company guide, sources and limits →Company evidence guideDiraq: from fabricated devices to usable qubitsRead the eight-qubit foundry result through the denominators that manufacturing requires.
It can show that a particular process supports functioning quantum devices under stated conditions. The next questions concern reproducibility and integration; a small result need not answer those questions to be technically useful.
Full company guide, sources and limits →Company evidence guideQuantum Motion: what makes a system installation reproducible?Turn a full-stack deployment claim into a record of what users can reproduce.
It can clarify what was delivered, accepted and made available within the testbed. The supplier and customer may describe different boundaries, so retaining both documents helps readers understand the actual service or research capability.
Full company guide, sources and limits →Company evidence guideOxford Ionics: how to read its benchmark historyKeep original experiments, acquisition history and later integration results in separate dated records.
No. It changes the organizational context, not the experiment that was performed. The result should retain its original provenance and be linked to the current ownership timeline.
Full company guide, sources and limits →Company evidence guideQblox: what does timing mean in a control system?Evaluate control electronics against an experiment’s signals and timing obligations.
The channels must support the required amplitude, noise, bandwidth, synchronization and readout relationships. A larger count cannot answer those compatibility questions by itself.
Full company guide, sources and limits →Company evidence guideQuantum Machines: why do calibration results need software versions?Trace a calibration claim through its controller, software versions, fit and final validation.
It can alter waveform construction, fitting, scheduling, parameter selection or result handling. A version record helps separate those changes from physical-device changes and makes a later comparison more meaningful.
Full company guide, sources and limits →Company evidence guideBluefors: which parts of a quantum system need cooling?A cryogenic partnership becomes useful when the cooled subsystem and thermal load are explicit.
The temperature requirements belong to particular components and detectors, not to the word photonic as a whole. A system-level description should identify those components and their operating conditions.
Full company guide, sources and limits →Company evidence guideArqit: what does a quantum-safe product protect?Classify ARQQ through its security products and examine the trust boundary behind the label.
No. These security products are intended to address cryptographic threats, including future quantum threats. The business and technical evidence therefore centre on deployment, assurance and integration rather than processor benchmarks.
Full company guide, sources and limits →Company evidence guideSEALSQ: how can a product certification be checked?Follow each chip from a certification target to an exact, checkable assurance record.
No. It states planned milestones. Completion requires the relevant authoritative record and the exact product configuration to which it applies.
Full company guide, sources and limits →