56Evidence
not yet measuredWhat it measures
Wartezeit bis zum ersten ausgegebenen Token.
What it does not measure
Gesamtdauer langer Antworten, Qualitaet.
How the score is composed
Currency50
Discrimination50
Headroom50
Contamination safety100
Gaming resistance100
Methodological rigor50
Reproducibility20
Details
- Benchmaxxing risk
- low
- Difficulty
- low
- Public test set
- no
- Human-verified
- no
- LLM-judged
- no
- Skala
- 0–30 Sekunden (niedriger besser)
Hand-maintained — no automatic adapter yet.
Best measured models
No results ingested yet.