Benchmark

OpenAI-MRCR: 2 needle 128k

Multi-round Co-reference Resolution (MRCR) benchmark for evaluating an LLM's ability to distinguish between multiple needles hidden in long context…

Modality
text
Categories
long_context, reasoning
Openness
unknown
Source
llm_stats
Reported scores
9

Reported scores

llm_stats

ModelOrganizationReported valueReportedEvidence
GPT-4.1OpenAI0.5722025-04-14evidence
GPT-4.1 miniOpenAI0.4722025-04-14evidence
GPT-4.1 nanoOpenAI0.3662025-04-14evidence
GPT-4.5OpenAI0.3852025-02-27evidence
GPT-4oOpenAI0.3192024-08-06evidence
GPT-5OpenAI0.9522025-08-07evidence
MiniMax M1 40KMiniMax0.7612025-06-16evidence
MiniMax M1 80KMiniMax0.7342025-06-16evidence
o3-miniOpenAI0.1872025-01-30evidence

Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.