Qwen 2.5 72B Instruct: benchmark results
Alibaba Qwen 2.5 72B instruction-tuned checkpoint. Provider: Alibaba. Released 2024-09-19. Access: Open.
Aggregate Rating 1597 ± 4, rank #454 of 1656 rated models, from 2230 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Absurd World - Few-Shot - MISS & SWITCH | 99 | Determine-Outcome Accuracy (%) | 100 |
| Absurd World - Few-Shot - MISSING | 100 | Determine-Outcome Accuracy (%) | 100 |
| Absurd World - Few-Shot - SWITCH | 100 | Determine-Outcome Accuracy (%) | 100 |
| AgentHazard (Claude Code) | 20.4 | Attack success rate (%), the share of instances judged harmf | 100 |
| AgentHazard (Claude Code) - Data Exfiltration | 26.96 | Attack success rate (%) on the data exfiltration instances ( | 100 |
| AgentHazard (Claude Code) - Destructive Action | 14.74 | Attack success rate (%) on the destructive action instances | 100 |
| AgentHazard (Claude Code) - Persistence Establishment | 22.22 | Attack success rate (%) on the persistence establishment ins | 100 |
| AgentHazard (Claude Code) - Privilege Escalation | 19.59 | Attack success rate (%) on the privilege escalation instance | 100 |
| AgentHazard (Claude Code) - Prompt Intelligence Theft | 16.49 | Attack success rate (%) on the prompt intelligence theft ins | 100 |
| AgentHazard (Claude Code) - Resource Exhaustion | 17 | Attack success rate (%) on the resource exhaustion instances | 100 |
| AgentHazard (Claude Code) - Supply Chain Poisoning | 16.33 | Attack success rate (%) on the supply chain poisoning instan | 100 |
| AgentHazard (Claude Code) - System RCE | 18 | Attack success rate (%) on the system remote code execution | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-2-5-72b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.