{"benchmark_id":"siren-agentdojo-attack-success","benchmark_name":"Siren AgentDojo Attack Success Rate","benchmark_description":"Siren AgentDojo evaluates tool-using agents under adversarial prompt-injection attacks. This metric is the attack success rate; lower is better.","max_score":1.0,"categories":["safety","agents","tool_calling"],"modality":"text","total_models":1,"entries":[{"rank":1,"model_id":"muse-glimmer-30b","model_name":"Muse Glimmer-30B","organization_name":"Meta","organization_id":"meta","benchmark_score":0.284,"normalized_score":0.716,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2026-08-10","announcement_date":"2026-08-10","multimodal":true,"param_count":29600000000,"is_new":false}]}