LiveCodeBench v6

LiveCodeBench v6 is a continuously refreshed competitive programming benchmark: it collects problems from Codeforces, AtCoder and LeetCode by publication date, so the model can never memorise the answers from training data. v6 is the edition with the newest problem window (2025).

LiveCodeBench v6 measures hard reasoning in science, technology, engineering and math. In detail, it checks:

• competitive programming with correct solutions
• no contamination: only problems newer than the model's cutoff
• pass@1 on executable, test-based tasks
• algorithmic skill across contest platforms

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.

What is the LiveCodeBench v6 benchmark?

LiveCodeBench v6 is a continuously refreshed competitive programming benchmark: it collects problems from Codeforces, AtCoder and LeetCode by publication date, so the model can never memorise the answers from training data. v6 is the edition with the newest problem window (2025).

What does LiveCodeBench v6 test?

LiveCodeBench v6 mainly measures hard reasoning in science, technology, engineering and math. Concretely, it checks: competitive programming with correct solutions; no contamination: only problems newer than the model's cutoff; pass@1 on executable, test-based tasks.

How is the LiveCodeBench v6 score calculated?

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.

Where do I get LiveCodeBench v6 data and scores?

The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/livecodebench-v6/scores) or pull it directly from the Open LLM Radar API with any HTTP client.