reasoning benchmark
Natural2Code
NaturalCodeBench (NCB) is a challenging code benchmark designed to mirror the complexity and variety of real-world coding tasks. It comprises 402 high-quality problems in Python and Java, selected from natural user queries from online coding services, covering 6 different domains.
Updated Aug 7, 2026
Published models8
Registry coverage8
MetricScore
EvidenceB
Natural2Code leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | GO | 92.9% | 100.0% | 8 | C | |
| 2 | GO | 85.4% | 85.7% | 8 | C | |
| 3 | GO | 84.5% | 71.4% | 8 | C | |
| 4 | GO | 80.7% | 57.1% | 8 | C | |
| 5 | GO | 79.8% | 42.9% | 8 | C | |
| 6 | GO | 75.5% | 28.6% | 8 | C | |
| 7 | GO | 70.3% | 14.3% | 8 | C | |
| 8 | GO | 56.0% | 0.0% | 8 | C |
Score distribution
Top published rows on the benchmark's original scale.
Natural2Code highlights
The top published results on this benchmark's own scale.
What is Natural2Code?
Definition and scoring fields from the benchmark registry.
NaturalCodeBench (NCB) is a challenging code benchmark designed to mirror the complexity and variety of real-world coding tasks. It comprises 402 high-quality problems in Python and Java, selected from natural user queries from online coding services, covering 6 different domains.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- Natural2Code
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- natural2code|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about Natural2Code.
Which model scores highest on Natural2Code?
Gemini 2.0 Flash is currently ranked first with 92.9%.
What does Natural2Code measure?
NaturalCodeBench (NCB) is a challenging code benchmark designed to mirror the complexity and variety of real-world coding tasks. It comprises 402 high-quality problems in Python and Java, selected from natural user queries from online coding services, covering 6 different domains.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
8 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.