Coding events,
judged correctly.

Code Arena is a platform for running coding contests. Organisers can create questions with top-of-the-line AI models, and a sandboxed execution engine grades every submission against hidden test cases in seconds. Live leaderboards, per-question scoring styles and analytics are built in, alongside a free Learn track. Browse open events.

Colleges and clubs running hackathons, teachers setting timed classroom assessments, and companies screening candidates. Participants just register and start solving - there is nothing to install, no compiler to set up and no repository to clone. A browser is the only requirement. Create an account.

Our code execution engine supports more than 60 languages, including C, C++, Java, Python, JavaScript, Go, Rust and Kotlin. You write and run your solution in the in-browser editor, and the options available for a question are shown in its language picker.

Your code runs in a sandbox against the organiser's test cases, capturing stdout and stderr. Every test must produce the expected output to count as accepted. Public test cases are visible while you work; hidden ones only run on submission, so you cannot tune your code to them.

Each question uses one of three scoring styles, set by the organiser, and a single event can mix all three. Static gives every accepted solve full marks, depreciating rewards solving first, and time-based shrinks your score as your personal clock runs down. Your event score sums your best score per question.

If some tests pass on a static or time-based question, you are offered points in proportion - the style score times tests passed over total tests, floored at 1. You can decline, keep zero and retry for full marks. Depreciating questions are all-or-nothing and never offer it.

By total score first, then fewer total submissions, then faster total execution time. Two people on the same score are separated by who needed fewer attempts, and if that ties too, by whose code ran faster overall. Only your best score on each question counts.

On AI-prompt questions you describe the solution and an AI model writes the code, and your latest generation is graded against the same test cases. Some cap how many tokens you may spend prompting: the fewer you use, the higher your multiplier, from x1.0 down to x0.75 as you near the cap. A live counter tracks usage.

Yes. Every submission is analysed in the background by custom machine-learning models that estimate how much of it came from external AI tools. A flagged submission can have its points reduced. Signals are reviewed by the organiser, with the Event Coordinator and Faculty Panel making the final call.

Register an organisation account, and once it is approved you get an organiser console: build a question bank, set per-question scoring styles and time budgets, upload a participant roster or open public registration, and follow submissions, the leaderboard and analytics live while the event runs. Register an organisation.