reasoning benchmark
LiveCodeBench
LiveCodeBench is a holistic and contamination-free evaluation benchmark for large language models for code. It continuously collects new problems from programming contests (LeetCode, AtCoder, CodeForces) and evaluates four different scenarios: code generation, self-repair, code execution, and test output prediction. Problems are annotated with release dates to enable evaluation on unseen problems released after a model's training cutoff.
Updated Aug 7, 2026
Published models73
Registry coverage73
MetricScore
EvidenceC
LiveCodeBench leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
Score distribution
Top published rows on the benchmark's original scale.
LiveCodeBench highlights
The top published results on this benchmark's own scale.
What is LiveCodeBench?
Definition and scoring fields from the benchmark registry.
LiveCodeBench is a holistic and contamination-free evaluation benchmark for large language models for code. It continuously collects new problems from programming contests (LeetCode, AtCoder, CodeForces) and evaluates four different scenarios: code generation, self-repair, code execution, and test output prediction. Problems are annotated with release dates to enable evaluation on unseen problems released after a model's training cutoff.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level C.
- Family
- LiveCodeBench
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- livecodebench|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about LiveCodeBench.
Which model scores highest on LiveCodeBench?
DeepSeek-V4-Pro-Max is currently ranked first with 93.5%.
What does LiveCodeBench measure?
LiveCodeBench is a holistic and contamination-free evaluation benchmark for large language models for code. It continuously collects new problems from programming contests (LeetCode, AtCoder, CodeForces) and evaluates four different scenarios: code generation, self-repair, code execution, and test output prediction. Problems are annotated with release dates to enable evaluation on unseen problems released after a model's training cutoff.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
73 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.