openrouter-evalsscopedweight 0.5adapter 10 of 17

OpenRouter Evals.

The OpenRouter Evals adapter aggregates evaluation outcomes observed across OpenRouter traffic: pass@1 on structured tasks and win-rates in comparative settings. It is a scoped, reduced-weight adapter because third-party eval aggregates corroborate capability without dominating the composite.

Counts only inside its own registry or ecosystem.

What it measures.

Pass@1

First-attempt success on structured tasks.

Win rate

Comparative win-rate aggregates across traffic.

Coverage

Breadth of tasks behind the aggregates.

Component keys.

openrouter-evals.*

Component family covering eval aggregates.

When it counts.

Scoped — counts for models and agents serving traffic through OpenRouter.

When signals are missing.

Subjects outside OpenRouter compute without this component.

How it contributes.

{
  "component": "openrouter-evals.*",
  "scale": "0-100",
  "mode": "scoped",
  "weight": 0.5,
  "contribution": "component x 0.5 / total weight"
}

The adapter’s components are averaged into the composite as a weighted mean: every contributing adapter’s component is multiplied by its weight, summed, and divided by the total weight of contributing adapters. The result is published as trustScores.total with a config version and timestamp so anyone can recompute it.

Questions about OpenRouter Evals.

What does the OpenRouter Evals adapter measure?

It aggregates pass@1 success rates and comparative win-rates observed across OpenRouter traffic into the openrouter-evals.* component family.

Why is the weight only 0.5?

Aggregate evals are useful corroboration but vary with traffic mix, so the catalog gives them a reduced weight relative to transactional and verification signals.