OpenAI MRCR (2 needles)1M

retrieval of two matching turns in the original MRCR evaluation

GPT-4.1 by OpenAI holds the best OpenAI MRCR (2 needles) (1M) result among tracked models, at 46.3% (released Apr 14 2025). Higher scores are better on this benchmark.

1. GPT-4.1 (OpenAI) — 46.3%; 2. GPT-4.1 mini (OpenAI) — 33.3%; 3. GPT-4.1 nano (OpenAI) — 12%.

3 tracked models have a published OpenAI MRCR (2 needles) (1M) score. Scores are collected from published lab reports and benchmark sources, with recorded sources credited alongside the results. We do not re-run the evaluations.