{"benchmark_id":"frontier-bench-v0.1","benchmark_name":"Frontier-Bench v0.1","benchmark_description":"Frontier-Bench v0.1 evaluates agentic terminal coding. Anthropic reports results using the mini-SWE-agent harness and a GKE backend, measured as mean reward across five attempts per task.","max_score":1.0,"categories":["agents","code"],"modality":"text","total_models":1,"entries":[{"rank":1,"model_id":"claude-opus-5","model_name":"Claude Opus 5","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":0.433,"normalized_score":0.433,"verified":false,"self_reported":true,"provider_id":"anthropic","input_cost_per_million":5.0,"output_cost_per_million":25.0,"speed_rps":null,"context_window":1000000,"release_date":"2026-07-24","announcement_date":"2026-07-24","multimodal":true,"param_count":null,"is_new":false}]}