Benchmarks

Internal test results. Treat every number on this page as an unverified internal figure — of the three scripts referenced below, only the k6 one currently exists, and it lives in a private monorepo. Reproduce-it-yourself lands with the BASIS 1.0 work.

Read the latency figures carefully. The enforcement numbers below are measured in-process — a direct function call, no network, no HTTP, no serialization. Cognigate's published ~38 ms design target is a different measurement: a full HTTP round trip, authorize plus sign plus return. Both are honest about what they measure; they are not comparable, and neither is a substitute for measuring your own deployment.

Destruction Test — 1,000 Agents

1,000 agents (85% normal, 10% bad actors, 5% manipulators) running full governance lifecycles through 100 stress rounds. In-process, single node — these latencies exclude network, HTTP, and serialization, so they are not comparable to an end-to-end API figure.

6.6ms
Enforcement p99
0%
Error rate
100%
Bad actor denial
102.7MB
Peak heap
MetricValueTargetvs target
Registration throughput104,068 agents/sec> 100within
Enforcement p50 (in-process)0.1ms< 5mswithin
Enforcement p95 (in-process)1.1ms< 50mswithin
Enforcement p99 (in-process)6.6ms< 100mswithin
Proof emission19,884/sec> 50/secwithin
Bad actor denial rate100.0%> 50%within
Peak heap usage102.7MB< 512MBwithin
Error rate0.00%< 1%within
323,200 actions processed. Every recorded figure sits within its stated target — but “within target” is not a test verdict: the script that produced these numbers is not published, so nobody outside Vorion can currently re-run it.

k6 HTTP Benchmark — Cognigate API

50 VUs (Normal Load)

Global p9519.79ms
Throughput286 req/sec
Error rate0%

500 VUs (Stress)

Global p953.49s
Throughput231 req/sec
Error rate0%
500 VU test saturates a single Node.js process. System stayed correct (0% errors) but slow. Production architecture uses 3+ replicas behind Traefik with KEDA autoscaling.

Enterprise Soak v3 — 500 Agents, 500K Signals

8,740
signals/sec
5,873
proof writes/sec
28,061
concurrent writes/sec
No leaks
memory stable
6/6 tests passed. 21-minute sustained run. 18 MB growth over 30 cycles (stable).

Scaling Path — Measured Base, Projected Ceiling

Only the first row is measured. Everything below it is a projection from CPU headroom and horizontal-scaling assumptions, not an observed result.

ConfigurationExpected RPSBasis
Single node (local)230Measured
Single node (4 vCPU VM)1,000-2,000CPU headroom
3 replicas + Traefik3,000-6,000Linear scaling
K8s + KEDA (10 pods)10,000-20,000Horizontal
K8s + KEDA (50 pods) + Redis50,000+Target

How These Were Produced

The commands below record how the figures were generated internally. They arenot yet runnable by you — see the note underneath.

# Destruction test (1,000 agents)
npx tsx scripts/destruction-test.ts

# k6 HTTP benchmark (requires cognigate-api running)
k6 run k6/steady-state.js

# Enterprise soak v3
npx vitest run tests/battle/enterprise-soak-v3.test.ts

Honesty note: only the k6 script currently exists (in Vorion's private monorepo); the destruction-test and soak scripts referenced above are not in any repository today. Treat every number on this page as an unverified internal result until the scripts are published — planned alongside the BASIS 1.0 work.