long context benchmark
NIH/Multi-needle
Multi-needle in a haystack benchmark for evaluating long-context comprehension capabilities of language models by testing retrieval of multiple target pieces of information from extended documents
Updated Aug 7, 2026
Published models1
Registry coverage1
MetricScore
EvidenceB
NIH/Multi-needle leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | ME | 84.7% | 100.0% | 1 | C |
NIH/Multi-needle highlights
The top published results on this benchmark's own scale.
What is NIH/Multi-needle?
Definition and scoring fields from the benchmark registry.
Multi-needle in a haystack benchmark for evaluating long-context comprehension capabilities of language models by testing retrieval of multiple target pieces of information from extended documents
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- NIH/Multi-needle
- Modality
- text
- Primary category
- long context
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- nih-multi-needle|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about NIH/Multi-needle.
Which model scores highest on NIH/Multi-needle?
Llama 3.2 3B Instruct is currently ranked first with 84.7%.
What does NIH/Multi-needle measure?
Multi-needle in a haystack benchmark for evaluating long-context comprehension capabilities of language models by testing retrieval of multiple target pieces of information from extended documents
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
1 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.