OJBench

OJBench evaluates competitive programming and reasoning on olympiad-informatics style problems. The model has to write full programs that pass hidden judge tests — correctness and runtime limits both count.

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

• algorithmic problem solving with strict test sets
• runtime and memory limits, as in a real judge
• complete program implementations
• reasoning about edge cases and complexity

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 OJBench benchmark?

OJBench evaluates competitive programming and reasoning on olympiad-informatics style problems. The model has to write full programs that pass hidden judge tests — correctness and runtime limits both count.

What does OJBench test?

OJBench mainly measures hard reasoning in science, technology, engineering and math. Concretely, it checks: algorithmic problem solving with strict test sets; runtime and memory limits, as in a real judge; complete program implementations.

How is the OJBench score calculated?

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

Where do I get OJBench 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/ojbench/scores) or pull it directly from the Open LLM Radar API with any HTTP client.