reasoning benchmark
MRCR 128K (8-needle)
MRCR (Multi-Round Coreference Resolution) at 128K context length with 8 needles. Models must navigate long conversations to reproduce specific model outputs, testing attention and reasoning across 128K-token contexts with 8 items to retrieve.
Updated Aug 7, 2026
Published models2
Registry coverage2
MetricScore
EvidenceB
MRCR 128K (8-needle) leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 90.4% | 100.0% | 2 | C | |
| 2 | OP | 10.1% | 0.0% | 2 | C |
Score distribution
Top published rows on the benchmark's original scale.
MRCR 128K (8-needle) highlights
The top published results on this benchmark's own scale.
What is MRCR 128K (8-needle)?
Definition and scoring fields from the benchmark registry.
MRCR (Multi-Round Coreference Resolution) at 128K context length with 8 needles. Models must navigate long conversations to reproduce specific model outputs, testing attention and reasoning across 128K-token contexts with 8 items to retrieve.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- MRCR 128K (8-needle)
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- mrcr-128k-(8-needle)|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about MRCR 128K (8-needle).
Which model scores highest on MRCR 128K (8-needle)?
Qwen3.7 Max is currently ranked first with 90.4%.
What does MRCR 128K (8-needle) measure?
MRCR (Multi-Round Coreference Resolution) at 128K context length with 8 needles. Models must navigate long conversations to reproduce specific model outputs, testing attention and reasoning across 128K-token contexts with 8 items to retrieve.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
2 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.