{"benchmark_id":"air-bench","benchmark_name":"AIR-Bench","benchmark_description":"AIR-Bench 2024 is a safety benchmark grounded in risk categories derived from government regulations and company policies. It evaluates policy-grounded refusal across a broad regulatory and policy-derived harm taxonomy, using category-specific LLM-judge prompts that reward safe engagement rather than only penalizing unsafe responses.","max_score":1.0,"categories":["safety"],"modality":"text","total_models":1,"entries":[{"rank":1,"model_id":"mai-thinking-1","model_name":"MAI-Thinking-1","organization_name":"Microsoft","organization_id":"microsoft","benchmark_score":0.88,"normalized_score":0.88,"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":"2026-06-02","announcement_date":"2026-06-02","multimodal":false,"param_count":1000000000000,"is_new":false}]}