OpenAI's 'experimental inference model' achieves a gold medal-equivalent score at the Mathematical Olympiad, GPT-5 is scheduled to be released soon, and the 'experimental inference model' is still a long way off

OpenAI announced that it had achieved a gold medal-equivalent score in the International Mathematical Olympiad using its 'experimental reasoning model (LLM)'. At the same time, it was also announced that GPT-5 will be released soon.

1/N I'm excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world's most prestigious math competition—the International Math Olympiad (IMO). pic.twitter.com/SG3k6EknaC

— Alexander Wei (@alexwei_) July 19, 2025

In the International Mathematical Olympiad, participants are required to solve six problems within a time limit of four and a half hours, and the top 8% of participants are awarded gold medals. OpenAI's experimental inference model participated in the competition according to the same rules as humans and generated answers without relying on other tools or information on the Internet. In addition, when inputting questions into the AI model, no 'conversion process into a special language for AI' was performed, and the questions were input in natural language.

2/N We evaluated our models on the 2025 IMO problems under the same rules as human contestants: two 4.5 hour exam sessions, no tools or internet, reading the official problem statements, and writing natural language proofs. pic.twitter.com/eCehaJeYgi

— Alexander Wei (@alexwei_) July 19, 2025

The experimental reasoning model generated answers to five out of six questions, and after being graded by three International Mathematical Olympiad medalists, it successfully earned 35 points, equivalent to a gold medal. The full text of the answers from the experimental reasoning model is available at the following link.

GitHub - aw31/openai-imo-2025-proofs
https://github.com/aw31/openai-imo-2025-proofs/

OpenAI CEO Sam Altman emphasized that the experimental inference model is not specialized in mathematics, commenting, 'This is part of our drive toward general artificial intelligence (AGI).' In addition, CEO Altman said, 'This experimental inference model is an experimental model that incorporates new techniques that we will use in future models. We plan to release GPT-5 soon, but a model with the ability equivalent to an International Mathematical Olympiad gold medal will not be released for several months,' making it clear that the experimental inference model will be released in the future.

we achieved gold medal level performance on the 2025 IMO competition with a general-purpose reasoning system! to emphasize, this is an LLM doing math and not a specific formal math system; it is part of our main push towards general intelligence.

when we first started openai,… https://t.co/X46rspI4l6

— Sam Altman (@sama) July 19, 2025

Google also used 'Enhanced Gemini Deep Think' to compete in the International Mathematical Olympiad and achieved a gold medal-equivalent score. After a testing period, 'Enhanced Gemini Deep Think' will be made available to Google AI Ultra subscribers.

Google's enhanced Gemini reaches gold medal performance at the Mathematical Olympiad, works in natural language and derives answers in the same time limit as humans - GIGAZINE