OpenAI has achieved a groundbreaking milestone with its experimental large language model (LLM), which performed at gold medal level during the 2025 International Math Olympiad (IMO). This advancement underscores the significant progress in AI reasoning abilities, confirmed by CEO Sam Altman as the precursor to the upcoming release of GPT-5. Researcher Alexander Wei shared that the AI successfully solved five out of six problems under authentic exam conditions, scoring 35 out of 42 points. The model’s performance is a notable leap from previous benchmarks, demonstrating its capabilities in generating intricate, human-like proofs.
Despite this success, OpenAI has no immediate plans to release the IMO-level model, focusing instead on future developments in AI general reasoning. Altman emphasized the model’s role in broader AI research, marking a significant leap in machine learning. With GPT-5 on the horizon, OpenAI is poised to redefine the limits of AI reasoning abilities in the near future.
Source link
