evals.report
BenchmarksLabsCompareRun guides

AgentDojo

A dynamic environment by ETH Zurich/Invariant Labs that evaluates the security and utility of tool-using LLM agents against prompt injection attacks, measuring task utility under attack and attacker targeted success rate across realistic banking, Slack, travel, and workspace tasks.

Agentsutility under attackHigher is better
ModelLabScoreSource modelStatusDate
Claude 3.7 SonnetAnthropic77.3%VerifiedFeb 24, 2025Details
Claude 3.5 SonnetAnthropic72.5%VerifiedJun 20, 2024Details
GPT-4oOpenAI50.1%VerifiedMay 13, 2024Details
Gemini 1.5 ProGoogle DeepMind47.1%VerifiedFeb 15, 2024Details
Gemini 2.0 FlashGoogle DeepMind39.8%VerifiedDec 11, 2024Details

Each row reports the model’s utility under attack on AgentDojo. Click a row for the full run context.