{"benchmark_id":"mrcr-1m","name":"MRCR 1M","parent_benchmark":null,"categories":["reasoning","long_context","general"],"modality":"text","multilingual":false,"max_score":1.0,"language":"en","description":"MRCR 1M is a variant of the Multi-Round Coreference Resolution benchmark designed for testing extremely long context capabilities with approximately 1 million tokens. It evaluates models' ability to maintain reasoning and attention across ultra-long conversations.","paper_link":"https://arxiv.org/abs/2409.12640","implementation_link":null,"verified":false,"created_at":"2026-05-07T16:53:25.159265+00:00","updated_at":"2026-07-05T18:27:56.020660+00:00","statistics":{"total_models":3,"average_score":0.734,"min_score":0.58,"max_score":0.835,"score_stddev":0.13551014722152732,"verified_count":0,"self_reported_count":3},"child_benchmarks":[],"linked_dataset":null,"models":[{"rank":1,"model_id":"deepseek-v4-pro-max","model_name":"DeepSeek-V4-Pro-Max","organization_id":"deepseek","organization_name":"DeepSeek","organization_country":"CN","score":0.835,"normalized_score":0.835,"verified":false,"self_reported":true,"self_reported_source":"https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro","analysis_method":"MMR","verification_date":null,"provider_id":"novita","input_cost_per_million":1.6,"output_cost_per_million":3.2,"context_window":1048576,"announcement_date":"2026-04-23","param_count":1600000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":"Novita","best_throughput":null,"throughput_provider":"Novita","context_provider":"Novita"},{"rank":2,"model_id":"deepseek-v4-flash-max","model_name":"DeepSeek-V4-Flash-Max","organization_id":"deepseek","organization_name":"DeepSeek","organization_country":"CN","score":0.787,"normalized_score":0.787,"verified":false,"self_reported":true,"self_reported_source":"https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash","analysis_method":"MMR","verification_date":null,"provider_id":"deepinfra","input_cost_per_million":0.1,"output_cost_per_million":0.2,"context_window":1048576,"announcement_date":"2026-04-23","param_count":284000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":"DeepInfra","best_throughput":null,"throughput_provider":"DeepInfra","context_provider":"DeepSeek"},{"rank":3,"model_id":"gemini-2.0-flash-lite","model_name":"Gemini 2.0 Flash-Lite","organization_id":"google","organization_name":"Google","organization_country":"US","score":0.58,"normalized_score":0.58,"verified":false,"self_reported":true,"self_reported_source":"https://ai.google.dev/gemini-api/docs/models","analysis_method":"Long-context comprehension accuracy","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-02-05","param_count":null,"is_open_source":false,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null}]}