{"benchmark_id":"cybench","benchmark_name":"CyBench","benchmark_description":"CyBench is a suite of Capture-the-Flag (CTF) challenges measuring agentic cyber attack capabilities. It evaluates dual-use cybersecurity knowledge and measures the 'unguided success rate', where agents complete tasks end-to-end without guidance on appropriate subtasks.","max_score":1.0,"categories":["safety","agents","code"],"modality":"text","total_models":2,"entries":[{"rank":1,"model_id":"claude-mythos-preview","model_name":"Claude Mythos Preview","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":1.0,"normalized_score":1.0,"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":null,"announcement_date":"2026-04-07","multimodal":true,"param_count":null,"is_new":false},{"rank":2,"model_id":"grok-4.1-thinking-2025-11-17","model_name":"Grok-4.1 Thinking","organization_name":"xAI","organization_id":"xai","benchmark_score":0.39,"normalized_score":0.39,"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":"2025-11-17","announcement_date":"2025-11-17","multimodal":true,"param_count":null,"is_new":false}]}