{"benchmark_id":"swe-marathon","benchmark_name":"SWE-Marathon","benchmark_description":"SWE-Marathon is an ultra-long-horizon software engineering benchmark covering tasks such as building compilers, optimizing kernels, and developing production-grade services. It measures whether agents can sustain quality across extremely long engineering trajectories.","max_score":1.0,"categories":["agents","code"],"modality":"text","total_models":8,"entries":[{"rank":1,"model_id":"muse-wonder-full","model_name":"Muse Wonder Full","organization_name":"Meta","organization_id":"meta","benchmark_score":0.575,"normalized_score":0.575,"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":"2026-10-12","announcement_date":"2026-10-12","multimodal":true,"param_count":null,"is_new":true},{"rank":2,"model_id":"muse-wonder","model_name":"Muse Wonder","organization_name":"Meta","organization_id":"meta","benchmark_score":0.531,"normalized_score":0.531,"verified":false,"self_reported":true,"provider_id":"meta","input_cost_per_million":4.0,"output_cost_per_million":20.0,"speed_rps":null,"context_window":1048576,"release_date":"2026-10-12","announcement_date":"2026-10-12","multimodal":true,"param_count":null,"is_new":true},{"rank":3,"model_id":"grok-4.7","model_name":"Grok 4.7","organization_name":"xAI","organization_id":"xai","benchmark_score":0.46,"normalized_score":0.46,"verified":false,"self_reported":true,"provider_id":"xai","input_cost_per_million":2.0,"output_cost_per_million":6.0,"speed_rps":null,"context_window":500000,"release_date":"2026-09-21","announcement_date":"2026-09-21","multimodal":true,"param_count":null,"is_new":false},{"rank":4,"model_id":"glm-5.3","model_name":"GLM-5.3","organization_name":"Zhipu AI","organization_id":"zai-org","benchmark_score":0.425,"normalized_score":0.425,"verified":false,"self_reported":true,"provider_id":"deepinfra","input_cost_per_million":1.2,"output_cost_per_million":4.0,"speed_rps":null,"context_window":1048576,"release_date":"2026-08-14","announcement_date":"2026-08-14","multimodal":false,"param_count":753000000000,"is_new":false},{"rank":5,"model_id":"kimi-k3","model_name":"Kimi K3","organization_name":"Moonshot AI","organization_id":"moonshotai","benchmark_score":0.42,"normalized_score":0.42,"verified":false,"self_reported":true,"provider_id":"deepinfra","input_cost_per_million":2.85,"output_cost_per_million":14.25,"speed_rps":null,"context_window":1048576,"release_date":"2026-07-16","announcement_date":"2026-07-16","multimodal":true,"param_count":2800000000000,"is_new":false},{"rank":6,"model_id":"hy4-preview","model_name":"Hy4 preview","organization_name":"Tencent","organization_id":"tencent","benchmark_score":0.319,"normalized_score":0.319,"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":"2026-08-28","announcement_date":"2026-08-28","multimodal":false,"param_count":770000000000,"is_new":false},{"rank":7,"model_id":"grok-4.5","model_name":"Grok 4.5","organization_name":"xAI","organization_id":"xai","benchmark_score":0.29,"normalized_score":0.29,"verified":false,"self_reported":true,"provider_id":"xai","input_cost_per_million":2.0,"output_cost_per_million":6.0,"speed_rps":80.0,"context_window":500000,"release_date":"2026-07-16","announcement_date":"2026-07-16","multimodal":true,"param_count":null,"is_new":false},{"rank":8,"model_id":"glm-5.2","model_name":"GLM-5.2","organization_name":"Zhipu AI","organization_id":"zai-org","benchmark_score":0.13,"normalized_score":0.13,"verified":false,"self_reported":true,"provider_id":"deepinfra","input_cost_per_million":0.75,"output_cost_per_million":2.4,"speed_rps":null,"context_window":1048576,"release_date":"2026-06-16","announcement_date":"2026-06-16","multimodal":false,"param_count":753000000000,"is_new":false}]}