That "best reasoning model ever" (LOL) is already far more intelligent than you and all of the other people who are coping in response to the parent comment, in reasoning and logic in general. Your dull and unremarkable mind, and the dull and unremarkable thoughts it produces make you less worthy of being considered sentient than that very "best reasoning model ever", whose thoughts in basically any field will be more novel and enlightening than yours. If you must cope in public, then do so without pathetically disparaging superior intelligences to yours.

Btw, ChatGPT can't be in 6th place because the information it used to solve the problem comes from submissions of others present on the internet.

Competitive Programming isn’t defined by some ai model spitting solutions to problems. It’s not about shortcuts, cheats, or farming fake progress. cp is about the community itself, the people who struggle, learn, share, compete, and grow together. As long as there are members pushing themselves, helping others, and loving the challenge, cp will never die.

When the calculator was invented, many thought mathematicians would lose their role. In reality, mathematics only grew stronger.

Many chess engines can beat grandmasters — does that replace the role of chess players? No. In reality, chess players still exist. They actually use them to train and become stronger than previous generations.

After all, who would enjoy a show from a lifeless machine?

absolutely not.

I mean that's great but CF as a contest platform cannot survive this. Nobody will take contests where any clown with chatGPT pro ranks above LGM.

of course, this is all just for capability showcase, they probably spent 4-5 figures if not 6 to run this test

Pray nobody gets their hands on the actually capable gpt5

To not to overshadow the participants achievements.

Upd : For those downvoting I was the Moroccan deputy leader for the IOI 2025 this was their argument in the GA meeting fyi.

Because Openai disdains and has no need to do that. Openai can easily get 6th in IOI and soon the 1st. Openai don't need to prove anything, because openai was SO GREAT! The foolish human will never witness how GREAT openai was. Everyone knows how WELL openai acted in our dailys.

GPT regularly gets very high performances in Codeforces rounds, as in users using GPT have got Top 50 in 2 of the last 3 Div1 contests, which is roughly equivalent to like IOI 10th or so. So I think this is very legit considering it is their internal model.

They do use a LOT of compute during the competition. An unreasonable amount. Much more energy than a human brain... Perhaps 100x? So the LLM itself isn't really that intelligent by itself, it's just a more efficient brute-force. Same goes for their IMO results.

If humans don't get plugged in, they would perform better. Not true for the model either.

I found the following in the March IC meeting minutes:

"Regarding the AI participation option, we considered different formats and ITC very strongly insisted on the format of such competition being that any interested "AI contestant" should bring their own hardware onsite to the IOI, so that we work on accommodating and isolating them so that they only connect to the Contest Management System network and their hardware directly makes submissions during the contest under the same conditions as contestants. Thus, we started some interesting AI participation rules discussions within ISC during the March meeting and will continue working towards defining such rules in case we have AI participation in IOI 2025, and we are only considering this onsite "bring their own hardware" participation option as advised by ITC."

So unless something changed since March, either they would have brought their own hardware to Bolivia (which would surely imply some kind of supervision from the contest organizers), or it's completely unofficial.

If I understood what was said in the GA correctly, they did in fact bring their own hardware.

you are wrong , just check this years problems.

The questions were not at all standard

Why am I getting so many downvotes? Do people think I meant CSES style problems? Obviously not, I meant ICPC style (which are much more standard than codeforces problems)

div 1 as well if this result is accurate 💀

Nah man, it was around $100M for GPT-4 which is pretty old model w/o reasoning, for newer models costs are in (tens of) billions for both electricity and hardware.

I don't see what the omnipotence paradox has to do with anything.

I believe competitive programming won't die for a much simpler reason — it is ultimately a recreational activity, and we like doing it. Computers beat humans in games like chess and Go years ago, are they dead? We will adapt if necessary, but I see little risk of the sport dying out because of AI.