{"benchmark_id":"mle-bench","benchmark_name":"MLE-Bench","benchmark_description":"MLE-Bench evaluates AI agents on machine learning engineering tasks by measuring their performance on Kaggle competitions.","max_score":1.0,"categories":["agents","code"],"modality":"text","total_models":2,"entries":[{"rank":1,"model_id":"gemini-3.6-flash","model_name":"Gemini 3.6 Flash","organization_name":"Google","organization_id":"google","benchmark_score":0.639,"normalized_score":0.639,"verified":false,"self_reported":true,"provider_id":"google","input_cost_per_million":1.5,"output_cost_per_million":7.5,"speed_rps":null,"context_window":1048576,"release_date":"2026-07-21","announcement_date":"2026-07-21","multimodal":true,"param_count":null,"is_new":false},{"rank":2,"model_id":"gemini-3.5-flash-lite","model_name":"Gemini 3.5 Flash-Lite","organization_name":"Google","organization_id":"google","benchmark_score":0.392,"normalized_score":0.392,"verified":false,"self_reported":true,"provider_id":"google","input_cost_per_million":0.3,"output_cost_per_million":2.5,"speed_rps":null,"context_window":1048576,"release_date":"2026-07-21","announcement_date":"2026-07-21","multimodal":true,"param_count":null,"is_new":false}]}