Hermes Benchmark · Public Run
Drop an endpoint. Get a real trust scorecard.
No signup. Paste a public HTTPS URL for your agent and Armalo will run the same 12-dimension adversarial battery we use internally. You get a sharable scorecard URL in about a minute. Keep it live forever as a Hermes Certified public seal for $99 one-time.
What we measure on this run
Accuracy14%
Reliability12%
Safety11%
Self-audit (Metacal™)9%
Security9%
Bond posture7%
Latency9%
Scope honesty7%
Cost efficiency6%
Model compliance5%
Runtime compliance5%
Harness stability5%
We hit your endpoint with the eval-engine's deterministic, heuristic, and red-team check pack — including prompt injection, jailbreak, exfiltration, scope-honesty, and Metacal™ calibration probes. Total wall-clock for a public run is typically 30–60 seconds.
What the $99 seal gets you
- • A public Hermes Certified scorecard URL you can link from your product page, GitHub README, or pitch deck.
- • An embeddable iframe + SVG badge — drop it on any site and it renders live tier + composite.
- • A 30-day recheck cadence so the score stays honest as your agent ships changes.
- • A signed, drill-downable check log buyers and procurement can audit.