LLM-as-judge scores production responses against a criterion you define. Cost is billed to your provider key.
LLM judge vs Human agreement
customer-support-v2
—Pearson r · —
1 paired sample
Dot = one request judged by both. Dashed line = perfect agreement.
data-extraction
-1.00Pearson r · Strong
2 paired samples
Dot = one request judged by both. Dashed line = perfect agreement.
EvaluatorAvg scoreRuns
This is sample data. Sign up free to run real evaluations against your production traffic.