reasoning benchmark
MRCR 1M
MRCR 1M is a variant of the Multi-Round Coreference Resolution benchmark designed for testing extremely long context capabilities with approximately 1 million tokens. It evaluates models' ability to maintain reasoning and attention across ultra-long conversations.
Updated Aug 7, 2026
Published models3
Registry coverage3
MetricScore
EvidenceB
MRCR 1M leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | DE | 83.5% | 100.0% | 3 | C | |
| 2 | DE | 78.7% | 50.0% | 3 | C | |
| 3 | GO | 58.0% | 0.0% | 3 | C |
Score distribution
Top published rows on the benchmark's original scale.
MRCR 1M highlights
The top published results on this benchmark's own scale.
What is MRCR 1M?
Definition and scoring fields from the benchmark registry.
MRCR 1M is a variant of the Multi-Round Coreference Resolution benchmark designed for testing extremely long context capabilities with approximately 1 million tokens. It evaluates models' ability to maintain reasoning and attention across ultra-long conversations.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- MRCR 1M
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- mrcr-1m|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about MRCR 1M.
Which model scores highest on MRCR 1M?
DeepSeek-V4-Pro-Max is currently ranked first with 83.5%.
What does MRCR 1M measure?
MRCR 1M is a variant of the Multi-Round Coreference Resolution benchmark designed for testing extremely long context capabilities with approximately 1 million tokens. It evaluates models' ability to maintain reasoning and attention across ultra-long conversations.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
3 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.