OpenAI's self - proclaimed Olympic math gold medal title has been severely criticized across the entire internet. Netizens angrily condemned it as hype, and math experts publicly raised doubts. Is it a claim to divinity or just a case of piggy - backing on others' fame?

Recently, a major event has stirred up a storm in the tech and math communities.

OpenAI announced that its AI model achieved a performance equivalent to a gold medal in the world's most prestigious math competition, the International Mathematical Olympiad (IMO).

In simple terms, AI has learned to solve IMO problems, and at a world - champion level.

IMO (International Mathematical Olympiad)

What Did OpenAI Do and Why the Controversy?

The first trigger of the storm is the "release time".

On July 19th, OpenAI researcher Alex Wei posted on the X platform that their experimental large - scale model achieved a gold - medal level performance in the IMO.

It got 5 out of 6 questions correct, scoring 35/42. The gold - medal standard is 35 points, and the silver - medal standard is 28 points.

The test was conducted in a standard exam environment without external tools, no internet access, with the same time limit and a single - submission rule.

The scoring was done by three former gold - medalists. Alex also posted the model's answer sheets on his GitHub.

Alex called it "a major leap in AI reasoning ability" and emphasized the model's generality. That is, the model used this time was not specially trained for the IMO. Its excellent performance comes from general reasoning ability. He specifically mentioned why this is a "big event":

Compared with past benchmarks, IMO problems require a new and continuous level of creative thinking. In terms of reasoning time, we've now moved from GSM8K (about 0.1 minutes for top humans) → MATH benchmark (about 1 minute) → AIME (about 10 minutes) → IMO (about 100 minutes).

However, the controversy started from the "release time" of Alex Wei's post about OpenAI achieving a gold - medal level performance in the IMO.

Hot post link: https://x.com/Mihonarium/status/1946880931723194389

Some netizens quoted the comment of IMO coordinator Joseph Myers,

At IMO 2025, the jury and coordinators generally considered it inappropriate for AI developers to announce IMO - related results before or after the IMO competition (e.g., before the closing party). Usually, coordinators believe that related announcements should be made at least one week after the closing ceremony because the focus of the competition should be on the achievements of human contestants, and reports about AI should not distract people.

Screenshot of Joseph Myers' comment, indicating that people he talked to associated with IMO all thought OpenAI's move was inappropriate.

Joseph Myers also said in the comment,

At IMO 2025, the jury and coordinators generally considered it inappropriate for AI developers to announce IMO - related results around the IMO competition time (e.g., before the closing party). Usually, the coordinators' view is that related announcements should be made at least one week after the closing ceremony because the focus of the competition should be on the achievements of human contestants, and reports about AI should not distract people.

Netizens also quoted a post from Harmonic, another math super - intelligent AI company participating in the IMO test, to confirm the IMO rule.

To respect the sanctity of the student competition, the IMO council has asked us and other participating major AI companies not to publish our results until July 28th.

OpenAI was thus pushed to the forefront of the storm of "disrespecting the competition". Netizens accused them of ruining the ceremony's sense of ritual for a headline, overshadowing the achievements of the young contestants under great pressure.

One netizen commented, "The young human contestants in IMO 2025 completed their answers under extremely high pressure, faced real reviewers, and took on the honor of rankings. Their achievements should be fully respected, instead of being overshadowed by OpenAI."

However, OpenAI researcher Noam Brown later clarified,

1. We released the news after the closing ceremony. The ceremony was live - streamed, which can prove our release time. 2. Before posting, we informed an IMO organizer in advance. The organizer requested us to wait until the ceremony ended, and we did follow this request.

Nevertheless, the labels of "jumping the gun" and "inappropriate behavior" still stuck to OpenAI. If this was just a misunderstanding in procedure and communication, the following controversy goes straight to the credibility of this "gold medal".

Whose Stage? Whose Rules? Who Decides?

The bigger question is, who certified this "gold medal"?

Although Alex Wei in his post, and later when OpenAI retweeted, they carefully stated that the model achieved a "gold medal - level performance", not "won a gold medal" or "won the competition". But in social - media communication, it was still simplified to "AI won the gold medal".

Since the IMO specifically invites relevant AI companies to participate in the competition, such as the aforementioned Harmonic and other AI companies, but OpenAI is not on the officially - invited list.

That is to say, OpenAI did not officially register for the competition, nor did it submit the results to the IMO review agency for scoring. Instead, it invited three former gold - medalists to "grade" the answers.

Some comments pointed out that the three former gold - medalists are not official coordinators. OpenAI's test process was not public. It was an internal verification and did not receive third - party verification like other AI challenges (such as the program battle with Psyho a few days ago).

It's more like an internal test rather than a public challenge. Although the model might have solved the problems, the non - public and non - transparent process made the claim of "gold - medal level" seem like "self - praise".

Despite the controversy of "jumping the gun on release time" and "untrustworthy gold - medal result", many netizens still supported OpenAI, believing it was harmless and could attract more attention.

One netizen said, "OpenAI's post made me click in to see who got the gold medal this year. I never cared before."

Some AI engineers also said, "This didn't overshadow the students. Instead, it made more people interested in the IMO."

Terry Tao: AI Solving Problems ≠ Equal to Humans

So, did OpenAI "cross the line" just to show off its strength, ignoring the feelings of the "kids"?

Terry Tao's speech at the IMO closing ceremony

Terry Tao, the award - presenting guest at the IMO, also posted a long article on Mastodon. Although he didn't directly talk about the OpenAI controversy, he shared his view:

There is a wide gap in AI capabilities, which may span several orders of magnitude. It depends on the resources and support provided and how the results are reported.