reasoning benchmark
MRCR v2 (8-needle)
MRCR v2 (8-needle) is a variant of the Multi-Round Coreference Resolution benchmark that includes 8 needle items to retrieve from long contexts. This tests models' ability to simultaneously track and reason about multiple pieces of information across extended conversations.
Updated Aug 7, 2026
Published models21
Registry coverage21
MetricScore
EvidenceC
MRCR v2 (8-needle) leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | OP | 91.5% | 100.0% | 21 | C | |
| 2 | OP | 89.6% | 95.0% | 21 | C | |
| 3 | AN | 76.0% | 90.0% | 21 | C | |
| 4 | OP | 74.0% | 85.0% | 21 | C | |
| 5 | GO | 66.4% | 80.0% | 21 | C | |
| 6 | GO | 60.1% | 75.0% | 21 | C | |
| 7 | GO | 54.0% | 70.0% | 21 | C | |
| 8 | GO | 44.1% | 65.0% | 21 | C | |
| 9 | GO | 43.4% | 60.0% | 21 | C | |
| 10 | OP | 41.3% | 55.0% | 21 | C | |
| 11 | OP | 33.6% | 50.0% | 21 | C | |
| 12 | OP | 33.1% | 45.0% | 21 | C | |
| 13 | GO | 26.6% | 40.0% | 21 | C | |
| 14 | GO | 26.3% | 35.0% | 21 | C | |
| 15 | GO | 26.3% | 30.0% | 21 | C | |
| 16 | GO | 25.4% | 25.0% | 21 | C | |
| 17 | GO | 22.1% | 20.0% | 21 | C | |
| 18 | GO | 21.3% | 15.0% | 21 | C | |
| 19 | GO | 19.1% | 10.0% | 21 | C | |
| 20 | GO | 16.4% | 5.0% | 21 | C | |
| 21 | GO | 13.5% | 0.0% | 21 | C |
Score distribution
Top published rows on the benchmark's original scale.
MRCR v2 (8-needle) highlights
The top published results on this benchmark's own scale.
What is MRCR v2 (8-needle)?
Definition and scoring fields from the benchmark registry.
MRCR v2 (8-needle) is a variant of the Multi-Round Coreference Resolution benchmark that includes 8 needle items to retrieve from long contexts. This tests models' ability to simultaneously track and reason about multiple pieces of information across extended conversations.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level C.
- Family
- MRCR v2 (8-needle)
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- mrcr-v2-(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 v2 (8-needle).
Which model scores highest on MRCR v2 (8-needle)?
GPT-5.6 Sol is currently ranked first with 91.5%.
What does MRCR v2 (8-needle) measure?
MRCR v2 (8-needle) is a variant of the Multi-Round Coreference Resolution benchmark that includes 8 needle items to retrieve from long contexts. This tests models' ability to simultaneously track and reason about multiple pieces of information across extended conversations.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
21 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is marked as eligible for the current LLMBoard capability methodology.