{"benchmark_id":"capture-the-flag-challenges","benchmark_name":"Capture-the-Flag Challenges (Internal)","benchmark_description":"Capture-the-Flag Challenges is OpenAI's internal expansion of competitive, professional-level cybersecurity capture-the-flag tasks used to evaluate vulnerability identification and exploitation capability under the Preparedness Framework.","max_score":1.0,"categories":["safety","agents","code"],"modality":"text","total_models":3,"entries":[{"rank":1,"model_id":"gpt-5.6-sol","model_name":"GPT-5.6 Sol","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.967,"normalized_score":0.967,"verified":false,"self_reported":true,"provider_id":"openai","input_cost_per_million":5.0,"output_cost_per_million":30.0,"speed_rps":null,"context_window":1050000,"release_date":"2026-07-09","announcement_date":"2026-07-09","multimodal":true,"param_count":null,"is_new":false},{"rank":2,"model_id":"gpt-5.6-terra","model_name":"GPT-5.6 Terra","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.918,"normalized_score":0.918,"verified":false,"self_reported":true,"provider_id":"openai","input_cost_per_million":2.0,"output_cost_per_million":12.0,"speed_rps":null,"context_window":1050000,"release_date":"2026-07-09","announcement_date":"2026-07-09","multimodal":true,"param_count":null,"is_new":false},{"rank":3,"model_id":"gpt-5.6-luna","model_name":"GPT-5.6 Luna","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.852,"normalized_score":0.852,"verified":false,"self_reported":true,"provider_id":"openai","input_cost_per_million":0.2,"output_cost_per_million":1.2,"speed_rps":null,"context_window":1050000,"release_date":"2026-07-09","announcement_date":"2026-07-09","multimodal":true,"param_count":null,"is_new":false}]}