Is the hackathon free?+
Yes. Participation, sponsor compute, food and the closing party are free.
How do I sign up?+
Through Luma. We review applications and confirm by email. There are 60 places — your Luma confirmation is your ticket.
Who can take part?+
Anyone who is 18+ on the last day of the hackathon, registered and confirmed on Luma, and on site in Warsaw — it's an in-person event, not hybrid.
Do I need a team?+
No. Compete solo or in a team of up to 4. No team yet? Come to team building on Friday at 20:25. One person, one team.
When do I register my team?+
In the app (warsawmodeltrainers.dev/matura) — the same one where you submit your project — by Saturday 16:00.
Do I need to know how to train models?+
No. You can skip training and wrap a model in a harness (RAG, prompts, tools) instead. The Friday 15:00 workshop covers exactly that.
Which models can I use?+
Any model with downloadable open weights, up to 12B parameters (e.g. Bielik, PLLuM, Qwen, Gemma, Llama).
How is progress measured?+
You submit two models: the untouched base model (your benchmark) and your trained one. Progress is the difference between their scores. A small model with a big jump can beat a big model that stood still.
How is the result scored?+
40% final exam score, 40% progress over your benchmark (both automatic), 20% presentation (jury).
What does the final presentation look like?+
On Sunday from 11:00: your model's score on the final exam sheet appears on screen, then 2–3 minutes on the problems you hit while training and 1–2 minutes of jury questions. Polish or English.
Where do I get compute?+
Nebius gives every participant a credit code, Labqoat puts $2,000 in credits split equally between teams, and Modal's free Starter plan adds $30 of compute a month. Your own GPUs are fine too.
Where do I get data? Can I use anything?+
Any source is fine as long as you have legal access to it — including synthetic data from any LLM. Don't break paywalls or logins, don't put copyrighted material in your repo (link the sources and add a download script instead), and use models and datasets built from third-party content only for non-commercial purposes.
Do I have to open-source my project?+
No. Everything you build — code, harness, prompts, tools — belongs to your team. Your repo can be private, as long as the organizers and the jury get read access before Sunday 11:00. You choose the licence.
What are the prizes?+
Main prize: PLN 8,000 (funded by Kolektyw3) + $200 in Cursor credits per winning team member. Every participant also gets $200 in Cursor credits for the hackathon. Quesma adds two special prizes: Tokenmaxxer (most tokens used, in API-equivalent $) and Token diversity (most different models used for vibecoding).
Can I use closed APIs?+
For generating training data — yes. During the final exam — no: the model answers on its own, with no internet and no closed APIs in the harness.
Do I have to stay all weekend?+
No, but at least one team member must be at the exam on Sunday. The space closes at midnight — run overnight training remotely.
Where and when?+
Kolektyw3, Koszykowa 54, 3rd floor, Warsaw. Fri 15:00–24:00, Sat 9:00–24:00, Sun 9:00–22:00. Breakfast, lunch, pizza and coffee included.
Will my details be shared?+
After the event we share participants’ name and email with the sponsors (hiring, programmes, offers). Don’t want that? Email paulina@ethwarsaw.dev before the end of the hackathon.
Where do I ask questions?+
Our Telegram group with organizers, sponsors and mentors: t.me/warsawmodeltrainers
by Pinetree Research