long context benchmark
OpenAI-MRCR: 2 needle 1M
Multi-Round Co-reference Resolution benchmark that tests an LLM's ability to distinguish between multiple similar needles hidden in long conversations. Models must reproduce specific instances of content (e.g., 'Return the 2nd poem about tapirs') from multi-turn synthetic conversations, requiring reasoning about context, ordering, and subtle differences between similar outputs.
Updated Aug 11, 2026
Models5
Model coverage5
MetricScore
EvidenceB
OpenAI-MRCR: 2 needle 1M Ranking
Higher score ranks better on this benchmark.
| 01 | MI | 58.6% | 100.0% | 5 | C | |
| 02 | MI | 56.2% | 75.0% | 5 | C | |
| 03 | OP | 46.3% | 50.0% | 5 | C | |
| 04 | OP | 33.3% | 25.0% | 5 | C | |
| 05 | OP | 12.0% | 0.0% | 5 | C |
OpenAI-MRCR: 2 needle 1M Score Distribution
A closer view of the leading scores on this benchmark.
OpenAI-MRCR: 2 needle 1M Highlights
The leading models and scores on this benchmark.
What is OpenAI-MRCR: 2 needle 1M?
What OpenAI-MRCR: 2 needle 1M measures and how its scores work.
Multi-Round Co-reference Resolution benchmark that tests an LLM's ability to distinguish between multiple similar needles hidden in long conversations. Models must reproduce specific instances of content (e.g., 'Return the 2nd poem about tapirs') from multi-turn synthetic conversations, requiring reasoning about context, ordering, and subtle differences between similar outputs.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- OpenAI-MRCR: 2 needle 1M
- Modality
- text
- Primary category
- long context
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- openai-mrcr:-2-needle-1m|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about OpenAI-MRCR: 2 needle 1M.
Which model scores highest on OpenAI-MRCR: 2 needle 1M?
MiniMax M1 40K is currently ranked first with 58.6%.
What does OpenAI-MRCR: 2 needle 1M measure?
Multi-Round Co-reference Resolution benchmark that tests an LLM's ability to distinguish between multiple similar needles hidden in long conversations. Models must reproduce specific instances of content (e.g., 'Return the 2nd poem about tapirs') from multi-turn synthetic conversations, requiring reasoning about context, ordering, and subtle differences between similar outputs.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
5 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.