← All benchmarks

GDPval - AgentCompass / GLM-5.3 Elo

Professional Work

AgentCompass/OpenHands evaluation on GDPval's 220 professional tasks with GLM-5.3 pairwise rubric judgments. Elo is computed within the reported comparison cohort and is kept distinct from Artificial Analysis's anchored GDPval-AA rating scales.

Top models (higher is better)