reasoning 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. 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.
Updated Aug 7, 2026
Published models9
Registry coverage9
MetricScore
EvidenceB
OpenAI-MRCR: 2 needle 128k leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | OP | 95.2% | 100.0% | 9 | C | |
| 2 | MI | 76.1% | 87.5% | 9 | C | |
| 3 | MI | 73.4% | 75.0% | 9 | C | |
| 4 | OP | 57.2% | 62.5% | 9 | C | |
| 5 | OP | 47.2% | 50.0% | 9 | C | |
| 6 | OP | 38.5% | 37.5% | 9 | C | |
| 7 | OP | 36.6% | 25.0% | 9 | C | |
| 8 | OP | 31.9% | 12.5% | 9 | C | |
| 9 | OP | 18.7% | 0.0% | 9 | C |
Score distribution
Top published rows on the benchmark's original scale.
OpenAI-MRCR: 2 needle 128k highlights
The top published results on this benchmark's own scale.
What is OpenAI-MRCR: 2 needle 128k?
Definition and scoring fields from the benchmark registry.
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.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- OpenAI-MRCR: 2 needle 128k
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- openai-mrcr:-2-needle-128k|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about OpenAI-MRCR: 2 needle 128k.
Which model scores highest on OpenAI-MRCR: 2 needle 128k?
GPT-5 is currently ranked first with 95.2%.
What does OpenAI-MRCR: 2 needle 128k measure?
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.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
9 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.