{"benchmark_id":"openai-mrcr:-2-needle-128k","name":"OpenAI-MRCR: 2 needle 128k","parent_benchmark":null,"categories":["reasoning","long_context"],"modality":"text","multilingual":false,"max_score":1.0,"language":"en","description":"Multi-round Co-reference Resolution (MRCR) benchmark for evaluating an LLM's ability to distinguish between multiple needles hidden in long context. Models are given a long, multi-turn synthetic conversation and must retrieve a specific instance of a repeated request, requiring reasoning and disambiguation skills beyond simple retrieval.","paper_link":"https://arxiv.org/abs/2403.05530","implementation_link":null,"verified":false,"created_at":"2026-05-07T16:53:25.609483+00:00","updated_at":"2026-07-05T18:27:57.999905+00:00","statistics":{"total_models":9,"average_score":0.5275555555555556,"min_score":0.187,"max_score":0.952,"score_stddev":0.24703902075942935,"verified_count":0,"self_reported_count":9},"child_benchmarks":[],"linked_dataset":null,"models":[{"rank":1,"model_id":"gpt-5-2025-08-07","model_name":"GPT-5","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.952,"normalized_score":0.952,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/index/introducing-gpt-5-for-developers/","analysis_method":"OpenAI-MRCR 2-needle retrieval at 128k tokens.","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-08-07","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},{"rank":2,"model_id":"minimax-m1-40k","model_name":"MiniMax M1 40K","organization_id":"minimax","organization_name":"MiniMax","organization_country":"CN","score":0.761,"normalized_score":0.761,"verified":false,"self_reported":true,"self_reported_source":null,"analysis_method":null,"verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-06-16","param_count":456000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":3,"model_id":"minimax-m1-80k","model_name":"MiniMax M1 80K","organization_id":"minimax","organization_name":"MiniMax","organization_country":"CN","score":0.734,"normalized_score":0.734,"verified":false,"self_reported":true,"self_reported_source":null,"analysis_method":null,"verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-06-16","param_count":456000000000,"is_open_source":true,"is_new":false,"best_latency":null,"latency_provider":null,"best_throughput":null,"throughput_provider":null,"context_provider":null},{"rank":4,"model_id":"gpt-4.1-2025-04-14","model_name":"GPT-4.1","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.572,"normalized_score":0.572,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/index/introducing-gpt-5-for-developers/","analysis_method":"Internal benchmark","verification_date":null,"provider_id":"openai","input_cost_per_million":2.0,"output_cost_per_million":8.0,"context_window":1047576,"announcement_date":"2025-04-14","param_count":null,"is_open_source":false,"is_new":false,"best_latency":10.0,"latency_provider":"OpenAI","best_throughput":100.0,"throughput_provider":"OpenAI","context_provider":"OpenAI"},{"rank":5,"model_id":"gpt-4.1-mini-2025-04-14","model_name":"GPT-4.1 mini","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.472,"normalized_score":0.472,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/index/introducing-gpt-5-for-developers/","analysis_method":"Internal benchmark","verification_date":null,"provider_id":"openai","input_cost_per_million":0.4,"output_cost_per_million":1.6,"context_window":1047576,"announcement_date":"2025-04-14","param_count":null,"is_open_source":false,"is_new":false,"best_latency":5.0,"latency_provider":"OpenAI","best_throughput":150.0,"throughput_provider":"OpenAI","context_provider":"OpenAI"},{"rank":6,"model_id":"gpt-4.5","model_name":"GPT-4.5","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.385,"normalized_score":0.385,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/index/gpt-4-1/","analysis_method":"Accuracy","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-02-27","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},{"rank":7,"model_id":"gpt-4.1-nano-2025-04-14","model_name":"GPT-4.1 nano","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.366,"normalized_score":0.366,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/index/gpt-4-1/","analysis_method":"Internal benchmark","verification_date":null,"provider_id":"openai","input_cost_per_million":0.1,"output_cost_per_million":0.4,"context_window":1047576,"announcement_date":"2025-04-14","param_count":null,"is_open_source":false,"is_new":false,"best_latency":2.0,"latency_provider":"OpenAI","best_throughput":200.0,"throughput_provider":"OpenAI","context_provider":"OpenAI"},{"rank":8,"model_id":"gpt-4o-2024-08-06","model_name":"GPT-4o","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.319,"normalized_score":0.319,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/index/gpt-4-1/","analysis_method":"Accuracy","verification_date":null,"provider_id":"openai","input_cost_per_million":2.5,"output_cost_per_million":10.0,"context_window":128000,"announcement_date":"2024-08-06","param_count":null,"is_open_source":false,"is_new":false,"best_latency":0.5,"latency_provider":"OpenAI","best_throughput":132.0,"throughput_provider":"OpenAI","context_provider":"OpenAI"},{"rank":9,"model_id":"o3-mini","model_name":"o3-mini","organization_id":"openai","organization_name":"OpenAI","organization_country":"US","score":0.187,"normalized_score":0.187,"verified":false,"self_reported":true,"self_reported_source":"https://openai.com/index/gpt-4-1/","analysis_method":"benchmark score","verification_date":null,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"context_window":null,"announcement_date":"2025-01-30","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}]}