reasoning benchmark
FrontierScience Research
FrontierScience Research is a benchmark evaluating AI models on cutting-edge scientific research questions requiring deep domain expertise, multi-step reasoning, and synthesis of complex scientific concepts across disciplines.
Updated Aug 7, 2026
Published models4
Registry coverage4
MetricScore
EvidenceB
FrontierScience Research leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | ME | 38.3% | 100.0% | 4 | C | |
| 2 | BY | 33.3% | 66.7% | 4 | C | |
| 3 | BY | 28.3% | 33.3% | 4 | C | |
| 4 | TE | 21.3% | 0.0% | 4 | C |
Score distribution
Top published rows on the benchmark's original scale.
FrontierScience Research highlights
The top published results on this benchmark's own scale.
What is FrontierScience Research?
Definition and scoring fields from the benchmark registry.
FrontierScience Research is a benchmark evaluating AI models on cutting-edge scientific research questions requiring deep domain expertise, multi-step reasoning, and synthesis of complex scientific concepts across disciplines.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- FrontierScience Research
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- frontierscience-research|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about FrontierScience Research.
Which model scores highest on FrontierScience Research?
Muse Spark is currently ranked first with 38.3%.
What does FrontierScience Research measure?
FrontierScience Research is a benchmark evaluating AI models on cutting-edge scientific research questions requiring deep domain expertise, multi-step reasoning, and synthesis of complex scientific concepts across disciplines.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
4 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.