QRF is a large-scale operational decision platform built around a class-independent core. The core operates on the decision data rather than being tied to one problem class; Routing Decision, AI Resource Allocation, and Storage Placement & Replication are the current independently validated examples, not the architectural boundary of QRF. Published engineering evidence reaches 500M nodes / ~3.0B edges in Routing and 100M scale in AI Resource and Storage on a standard desktop-class PC. No GPU, server, workstation, or quantum hardware is required for the published evidence.
What matters: Class-Independent Core · Massive Scale · Low Latency · Standard Hardware
Published QRF engineering benchmarks were produced on a standard consumer desktop PC rather than a server, workstation, GPU platform, or specialized HPC system.
Benchmark hardware disclosure is included to make the scale-to-hardware relationship explicit.
QRF separates the decision core from the operational class. The published classes demonstrate the same core across distinct decision problems while keeping each class benchmarked independently with its own workload, result type, and measurement scope.
Current evidence: Routing Decision, AI Resource Allocation, and Storage Placement & Replication are independently validated. Additional decision domains may be evaluated separately as QRF expands.
QRF is built around one decision core rather than a separate engine for each problem class. The published evidence shows that core applied to distinct operational decisions while preserving class-specific workloads, results, and measurement scope. The result is a platform proposition defined by scale, latency, hardware efficiency, and class independence—not by a single benchmark or use case.
QRF Routing evidence spans from 1M to 500M tested nodes and reaches approximately 3.0B directed edges. The objective is to demonstrate that decision-serving remains practical as graph size increases, rather than proving performance only on small or laboratory-scale datasets.
At the tested 100M scale, AI Resource Allocation and Storage Placement & Replication remain below 1 ms median p95 in the native decision pipeline. This matters because the decision workload grows with system scale, yet the measured serving latency remains suitable for low-latency technical evaluation.
Published engineering benchmarks were produced on a standard consumer desktop configuration using an Intel Core i5-12400F and 16 GB RAM. No GPU acceleration, professional workstation, server-class platform, quantum hardware, or specialized HPC environment was required for the published evidence.
QRF does not bind its decision core to Routing, AI Resource, Storage, or any other single operational class. The three published classes are independently validated examples of the same core operating across different decision problems; they define the current evidence base, not the architectural boundary.
Each operational class keeps its own workload, measurement scope, result type, latency figures, throughput figures, and validation evidence. Routing metrics are not reused as AI or Storage claims, which makes the published results easier to interpret and technically defend.
Organizations can begin with a controlled Technical Evaluation using an agreed scale, workload, constraints, and success criteria. Only after the evidence is reviewed does the engagement need to move toward a customer pilot or integration discussion, reducing technical and commercial commitment risk.
One core. Multiple decision classes. Class-specific evidence. QRF combines large-scale execution, low-latency decision serving, and standard-hardware operation without treating one benchmark as proof for unrelated workloads. Technical teams can evaluate the core against their own decision class, scale, constraints, and success criteria.
These are potential application domains for QRF operational classes. They are not themselves QRF classes.
Potential evaluation for vehicle-to-everything (V2X) and vehicular ad hoc network (VANET) scenarios where topology, connectivity, and routing conditions can change rapidly across connected vehicles and roadside infrastructure.
Potential evaluation for IoT/IIoT environments with large numbers of connected devices, gateways, and edge nodes where routing decisions must remain efficient as the network grows and conditions change.
Potential evaluation with automotive OEMs, Tier-1 suppliers and mobility platforms for large-scale connected-vehicle routing and communication scenarios under customer-defined technical and operational requirements.
Potential collaboration with telecom, networking and edge-computing teams evaluating routing-decision performance across distributed, large-scale, and changing network topologies.
Potential evaluation for Intelligent Transportation Systems (ITS), roadside infrastructure, and connected urban networks where large numbers of moving and fixed nodes interact.
QRF can also be evaluated in products that depend on large graph-based routing decisions and require an alternative to treating shortest path as the only routing objective.
QRF is available for technical evaluation and integration discussions with automotive OEMs, Tier-1 suppliers, semiconductor and networking companies, telecom vendors/operators, IoT and edge-platform providers, and smart-mobility / ITS teams.
Routing Decision, AI Resource Allocation, and Storage Placement & Replication are the current independently validated QRF operational classes. They are published examples of a class-independent decision core, not a statement that QRF is architecturally limited to those three domains. Additional decision classes can be evaluated separately and should be presented as validated only after comparable class-specific evidence exists.
The current three classes establish evidence across distinct decision problems while keeping the core conceptually separate from the class. They demonstrate breadth without turning untested domains into public claims. New classes should follow the same evidence-first path: technical evaluation, controlled measurement, result review, and only then public validation.
Routing Decision is QRF Class 01. The routing evidence below remains specific to this class and is not used as an AI Resource or Storage KPI.
Watch QRF execute and validate routing queries on a 100M-node / 600M-edge graph.
Controlled technical demonstration. Demo measurements are illustrative and are not an end-to-end production KPI or SLA.
| Scale | Edges | Scale Class | Memory Mode | Core p95 | Stable Runs |
|---|---|---|---|---|---|
| 1M | 6M | Standard-scale | Memory-resident | 0.233 ms | 5/5 |
| 10M | 60M | Standard-scale | Memory-resident | 0.441 ms | 5/5 |
| 100M | 600M | Standard-scale | Memory-resident | 1.025 ms | 5/5 |
| 200M | 1.2B | Large-scale | Memory-adaptive | 2.795 ms | 5/5 |
| 500M | 3.0B | Large-scale | Memory-adaptive | 6.997 ms | 5/5 |
The figures above report QRF Core Request-Serving p95 benchmark results. This metric is not an end-to-end customer-production KPI or SLA.
Published executive benchmark values
Published executive benchmark values
Published results span 1M to 500M tested nodes, approximately 6M to 3.0B edges, and 25/25 stable benchmark runs.
An integration-ready QRF Web Service is available for customer-scoped integration following Technical Evaluation.
Organizations can request a Technical Evaluation to assess QRF against their operational decision problem, scale, constraints, and evaluation objectives.
If technical fit is established, QRF can move into a scoped Customer Pilot through its integration-ready Web Service.
Technical Review → Technical Evaluation → Customer Pilot.
AI Resource Allocation is QRF Class 02. It evaluates AI-infrastructure resource decisions and returns class-specific allocation results. Its benchmark evidence is independent from Routing.
| Scale | Workload | Mean / request | p95 | Throughput | Stable Runs |
|---|---|---|---|---|---|
| 1M | 1000 frozen jobs/run | 0.200 ms | 0.223 ms | 4,662 jobs/s | 5/5 |
| 10M | 1000 frozen jobs/run | 0.284 ms | 0.310 ms | 3,344 jobs/s | 5/5 |
| 100M | 1000 frozen jobs/run | 0.369 ms | 0.432 ms | 2,617 jobs/s | 5/5 |
AI Resource metrics report the native class-specific decision pipeline; they are not Routing/Core p95 values and are not an SLA.
Median native decision-pipeline p95 across five runs per scale
Storage Placement & Replication is QRF Class 03. It evaluates distributed-storage placement and replication decisions and returns a class-specific storage result. Its evidence is independent from Routing and AI Resource Allocation.
| Scale | Workload | Mean / request | p95 | Throughput | Stable Runs |
|---|---|---|---|---|---|
| 1M | 1000 frozen requests/run | 0.200 ms | 0.217 ms | 4,603 req/s | 5/5 |
| 10M | 1000 frozen requests/run | 0.287 ms | 0.305 ms | 3,293 req/s | 5/5 |
| 100M | 1000 frozen requests/run | 0.371 ms | 0.403 ms | 2,585 req/s | 5/5 |
Storage latency and throughput are class-specific engineering measurements. Placement-success/quality evidence should be published separately from latency evidence.
Median native decision-pipeline p95 across five runs per scale
Evaluate QRF against your decision problem. Request a Technical Evaluation to define the operational class, scale, constraints, success criteria, and potential customer integration.
A focused path for CTO and engineering teams to validate QRF against their own requirements without committing to a production integration.
Published benchmark figures are engineering measurements within the stated Core Request-Serving scope and are not end-to-end customer-production KPI/SLA guarantees.
Benchmark hardware: standard desktop-class PC with Intel Core i5-12400F CPU, 16 GB RAM, and Windows 11 Pro. No GPU acceleration or specialized server/HPC hardware was used.
