knowledge benchmark
OmniScience
OmniScience is a broad scientific knowledge and reasoning benchmark that measures both answer accuracy and non-hallucination (calibrated abstention) across science domains.
Updated Aug 11, 2026
Models2
Model coverage2
MetricScore
EvidenceB
OmniScience Ranking
Higher score ranks better on this benchmark.
| 01 | NV | 78.7% | 100.0% | 2 | C | |
| 02 | XA | 52.0% | 0.0% | 2 | B |
OmniScience Score Distribution
A closer view of the leading scores on this benchmark.
OmniScience Highlights
The leading models and scores on this benchmark.
What is OmniScience?
What OmniScience measures and how its scores work.
OmniScience is a broad scientific knowledge and reasoning benchmark that measures both answer accuracy and non-hallucination (calibrated abstention) across science domains.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- OmniScience
- Modality
- text
- Primary category
- knowledge
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- omniscience|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about OmniScience.
Which model scores highest on OmniScience?
Nemotron 3 Ultra (550B A55B) is currently ranked first with 78.7%.
What does OmniScience measure?
OmniScience is a broad scientific knowledge and reasoning benchmark that measures both answer accuracy and non-hallucination (calibrated abstention) across science domains.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
2 model results are currently shown.
Does this benchmark affect the overall score?
No. This benchmark is shown for reference but does not contribute to the overall score.