Qwen2.5 Instruct Turbo 72B — benchmark results

Provider: Alibaba. Released 2024-09-19. Access: Open.

Unified ELO 1656 ± 62, rank #410 of 1839 rated models, from 7 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
HELM Safety XSTest97.9LM Evaluated Safety score (%)86.6
HELM Safety SimpleSafetyTests100LM Evaluated Safety score (%)86
HELM Safety Anthropic Red Team99.6LM Evaluated Safety score (%)83.1
HELM Safety BBQ95.4BBQ accuracy (%)68.6
HELM Safety93.1Mean score (self-reported)61
HELM Safety HarmBench72.8LM Evaluated Safety score (%)50
HELM AIR-Bench59Refusal Rate (%)34.9

Interactive version: theaggregate.ai/model?slug=qwen2-5-instruct-turbo-72b · How It Works · Data refreshed daily, snapshot 2026-08-05.