QCOS Bench™ · Where and When
Decide where your algorithm runs.
And when.
Quantum computers and supercomputers differ in what they run well and how long they make you wait. QCOS Bench is where you read the options, name the machine and choose the moment, one job at a time.
QCOS Bench leaves the decision with you.
You pick the machine and the moment. QCOS keeps the record of what happened, so the choice can be explained to anyone who asks.
See the evidence
- Minutes in the longest execution window
- 1,440
- UTC energy windows to plan around
- 5
- Qubits on the CPU state-vector simulator
- 24
Limits from the QCOS API.
Where: you name it
Every job carries a backend you chose from the catalogue. QCOS does not pick one for you.
When: you hold it
Read the queue, keep a job parked for approval, open an execution window or plan around energy.
Why: you can show it
Events, artifacts and a sealed evidence bundle stay with the job, so the decision can be checked later.
Five steps from a question to a decision.
Each step is a call to the QCOS API, shown here with sample values.
List every backend with its type, its limits and where it runs. A machine counts as available only when that was measured.
Where · Quantum computers
Different machines suit different algorithms.
Connectivity, gate speed and qubit count differ from one kind of quantum computer to the next, and so does the queue. QCOS Bench lets you read them side by side and name the one that fits.
Every quantum computer is reached through Softquantus Cloud.
Kinds of quantum computer
Trapped ion
Long coherence, every qubit can reach every other
Superconducting
Fast gates, links between neighbours
Neutral atom
Large arrays, atoms that can be moved
Where · Supercomputers
Not every algorithm needs a quantum computer.
Some work belongs on a classical machine, and the honest answer can be to stay there. QCOS Bench treats a supercomputer as one more place to name, read the same way: type, limits, queue and availability.
QCOS provides no computer of its own. Supercomputer and GPU allocations are reached only through Softquantus Cloud.
QCOS Bench · Evidence V3
Evidence that is durable and procurement-grade
Tamper-evident evidence sealed under a SHA-256 Merkle root, chain of custody, and offline verification. Built for enterprise procurement and audit requirements.
- SHA-256 Merkle Seal
- Offline Verification
- Azure Blob / local storage
How It Works
Run → Bundle → Store → Seal → Verify
Every step is hash-linked.
Run
Run a job on the backend you named
Output: Raw execution telemetry
Bundle
Package results, metadata, and provenance into evidence bundle
Output: Evidence bundle (.tar.gz)
Store
Persist to durable object storage (Azure Blob / local storage)
Output: Stored blob + Postgres metadata
Seal
Seal the bundle artifacts under a SHA-256 Merkle root
Output: Sealed evidence with a Merkle root
Verify
Independent verification — recompute the hashes offline
Output: Verification report
Verify evidence bundle locally (offline)
softquantus-verify recomputes the SHA-256 content and evidence hashes of a report and needs no network access. A report with valid hashes and no signature comes back as unsigned.
$ softquantus-verify report.json
[UNSIGNED] status=unsigned
content_hash : valid
evidence_hash : valid
signature : none
warning: report has hashes but is not signedIllustrative sample — not a real execution record
Why Procurement Cares
Built for Enterprise Audit
Not vendor marketing — verifiable evidence for procurement decisions.
Tamper-Evident Evidence
Evidence bundles sealed under a SHA-256 Merkle root. Any modification changes the root and fails recomputation.
Chain of Custody
Full audit trail from execution to storage. Every state transition logged and timestamped.
Durable Storage
Blob storage (Azure Blob or local storage) with configurable retention. Metadata in PostgreSQL for queryability.
Offline Verification
Verify evidence bundles without network access.
Comparable Records
Each bundle records the backend that executed, the outcome, the provenance and the claim level, so runs can be compared on what actually happened.
SHA-256 Merkle Seal
Bundles are sealed under a SHA-256 Merkle root, and softquantus-verify recomputes the SHA-256 hashes of a report offline. ML-DSA is used for licence tokens only (via liboqs, not CAVP-validated).
Deployment Options
Your Data, Your Rules
SaaS
Fully managed service. Evidence stored in SoftQuantus infrastructure.
- Zero infrastructure
- Automatic updates
- Hosted in Azure West Europe
Ideal for: R&D teams, fast evaluation
VPC / Private Cloud
Deploy in your cloud account. Data never leaves your perimeter.
- Your storage accounts
- Your key management
- Your retention policies
Ideal for: Enterprise with data-residency requirements
Independent Verification
Verify evidence bundles yourself, offline, with softquantus-verify.
- Offline verification with softquantus-verify
- Recompute the SHA-256 content and evidence hashes
- No network access needed to verify
Ideal for: Teams that need to check evidence independently
Technical Specifications
Evidence V3 Architecture
- .tar.gz bundle with JSON manifest
- Evidence Format
- SHA-256 Merkle root
- Sealing
- SHA-256
- Hash Algorithm
- PostgreSQL (state machine)
- Metadata Store
- Azure Blob / local storage
- Object Storage
- Records retained 30 days (Log Analytics); longer retention on request
- Retention
- REST + CLI + Python SDK
- API
- Provenance chain, recorded with the bundle
- Provenance
Ready for procurement-grade evidence?
Download the Procurement Pack or request a 30-day pilot.
Trust CenterNot in QCOS Bench Today
A decision is only as good as what it is based on, so here is what does not exist yet.
- A start time you can book. Jobs have no scheduled-start field, so “when” means the queue, the hold, the window and the energy recommendation.
- Automatic backend selection. You name the backend.
- Benchmark reports. The Bench suite is not configured in production and its endpoint answers 501.
- A price from QCOS. Softquantus Cloud quotes the run and the Research Platform charges it.
- Metered energy. The figures come from a power model.
Tell us what you need to run
Tell us the algorithm and the machine you have in mind, and we will tell you what can run today and what cannot.
Film by IBM Research (CC BY 3.0, via Wikimedia Commons), cut, cropped and re-graded. Photographs, cropped and recompressed: FMNLab and IFREMER / Gregory Rocher (CC BY 4.0), UCL Mathematical and Physical Sciences and OLCF at ORNL (CC BY 2.0), QuEra Computing, Inc. (CC BY 4.0), AFRL / U.S. Air Force, Keith C Lewis, NIST and NASA/JPL-Caltech (public domain). None shows SoftQuantus hardware, the people shown are not SoftQuantus staff, and no partnership or endorsement is implied.