Official task materials from the International Olympiad in Artificial Intelligence 2026, held in Astana, Kazakhstan.
The olympiad brought together 462 contestants from 103 delegations, across an at-home qualifying round and two on-site competition days.
| Folder | What's inside |
|---|---|
At-Home-Round |
Three preparatory notebook tasks, released before the olympiad |
Individual-Contest |
The six on-site tasks — statements, figures, and translations |
GAITE-Contest |
Hints for the GAITE track, which used the same six tasks |
Three open notebook tasks, worked on before arriving in Astana.
| # | Task | Proposed by | Developed by |
|---|---|---|---|
| 1 | Operation Night Watch | Anna Piunova | Anna Piunova · Magauiya Zhussip |
| 2 | Robot Delivery Academy | Salem Lahlou · Anuar Aimoldin | Anuar Aimoldin |
| 3 | The Analytical Language of John Wilkins | Nurdaulet Akhanov | Nurdaulet Akhanov |
The main competition: three tasks per day over two on-site days, each solved in a fresh notebook environment on one GPU (≈16 GB VRAM) with no internet access.
Day 1 — Find the Order · Robot Chasing · Potato Day 2 — Double Agent Dilemma · Ghost of the Machine · IOAI Field
Materials for IOAI Field are not published in this repository.
Each task folder holds the English statement in Markdown and PDF, its figures, and two sets of translations — the delegations' reviewed versions and a broader set of machine translations. See the Individual Contest README for task summaries, statement limits, the translation policy, and full credits.
GAITE is a simplified entry track for delegations that have recently joined IOAI. It used exactly the same six tasks as the Individual Contest, with three differences: per-task hints were provided, the LLM chatbot available to contestants was slightly more capable, and it was scored on its own ranking. Only the hints live in GAITE-Contest; all other task materials are shared with the Individual Contest.
Statements were translated into the delegations' languages so that contestants could work in their first language. Two sets are published for every Individual Contest task, and they are not interchangeable:
Translations/— the versions actually given to contestants, each read and corrected by that delegation's team leader. Authoritative.AI-Translation/— machine translations into 44 languages, generated automatically and not human-verified. Broader coverage, lower confidence.
Where the two disagree, Translations/ wins. The English statement at the root of each task folder is the source of truth for both.
The tasks, contest platform, and translations were the work of the Host and International Scientific Committees, the International Jury, the delegation team leaders, and the Yandex Contest Platform team. Everyone is credited by name in the Individual Contest README, as presented in the International Jury Report at the closing ceremony.
These materials are released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0). You are free to share and adapt them, including commercially, provided you give appropriate credit, link to the licence, and indicate any changes you made.