Next-Gen AI Model Achieves New Milestone in General Intelligence

Alexander Wei, a member of Openai’s Large Language Model (LLM) and reasoning research team, reported that a next-generation artificial intelligence (AI) experimental reasoning model achieved gold medal-level performance at the International Math Olympiad (IMO). Openai CEO Sam Altman emphasized that this is part of their effort to advance general intelligence.

Openai’s Experimental AI Model Receives Gold Medal at International Math Olympiad

Openai, a leading company in the field of artificial intelligence, has achieved a new milestone with a general-purpose model. An experimental reasoning model for general intelligence reached gold medal-level performance at the International Math Olympiad (IMO), one of the oldest and most prestigious mathematics competitions in the world.

Alexander Wei, a researcher in Openai’s reasoning team, reported that the experimental model faced the same conditions as any human taking the test, having no internet or tools to solve six problems in a span of two 4.5-hour sessions.

Openai’s experimental model successfully solved 5 of the 6 problems presented in the 2025 IMO, providing natural language proofs. Wei explained that after being evaluated by three former IMO medalists, the model earned 35 out of a total of 42 points, sufficient to achieve gold medal status.

Highlighting the relevance of this result, Wei stressed:

We reach this capability level not via narrow, task-specific methodology, but by breaking new ground in general-purpose reinforcement learning and test-time compute scaling.

Sam Altman also weighed in on the issue, emphasizing that the goal was reached using a general model and not one trained for math purposes. “It is part of our main push towards general intelligence,” he added.

Altman noted that this achievement is a “significant marker of how far AI has come over the past decade.”

Nonetheless, do not expect such an advanced reasoning model to be released for general audiences anytime soon. Altman reinforced that while version 5 of the company’s signature model, ChatGPT, will be released soon, it will not be able to reason in the same way as this yet unnamed experimental model. “We think you will love GPT-5, but we don’t plan to release a model with IMO gold level of capability for many months,” he concluded.

Read more: OpenAI Announces GPT-4.5: ‘A Giant, Expensive Model’