Sam Altman says OpenAI LLM achieved IMO gold-level Math skills, GPT-5 launch coming soon
Evaluated under same conditions as human participants
GPT-5 coming soon, but won't match IMO model's capabilities
An experimental large language model (LLM) developed by OpenAI has achieved gold medal-level performance at the 2025 International Math Olympiad (IMO), setting a new benchmark in mathematical reasoning for AI systems. Announcing the milestone, OpenAI researcher Alexander Wei posted on X that the model solved five out of six problems from the latest IMO under human exam conditions. The model earned 35 out of 42 possible points, a score that would qualify for a gold medal at the real competition.
'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,' Wei explained.
The IMO is regarded as the most prestigious high school maths competition globally, known for its notoriously complex problems. Wei pointed out that such problems demand extended creative reasoning and that achieving gold-level performance represents a leap from earlier benchmarks. 'We've now progressed from GSM8K (~0.1 min for top humans) MATH benchmark (~1 min) AIME (~10 mins) IMO (~100 mins),' he said.
Submissions were graded independently by three former IMO medallists, who unanimously validated the model's solutions. According to Wei, 'the model solved P1 through P5; it did not produce a solution for P6.' He shared the model's answers publicly, noting its 'distinct style,' owing to its experimental nature.
Wei said what makes the result even more impressive is that IMO proofs are long, complex and hard to verify. "By going beyond the reinforcement learning paradigm of clear-cut, verifiable rewards we've obtained a model that can craft intricate, watertight arguments at the level of human mathematicians.'
The LLM that achieved this result will not be released publicly any time soon. Wei clarified that while OpenAI is preparing to launch GPT-5, this IMO-level model is part of a different research track. 'We don't plan to release a model with IMO gold level of capability for many months.'
OpenAI CEO Sam Altman echoed this in a follow-up post, calling the achievement 'a significant marker of how far AI has come over the past decade.' He clarified that this model is not a specialised maths system, but a general-purpose reasoning model. 'We are releasing GPT-5 soon but want to set accurate expectations: this is an experimental model that incorporates new research techniques we don't plan to release a model with IMO gold level of capability for many months,' Altman added.
Looking back, Wei also reflected on how far AI progress has exceeded expectations. 'In 2021, my PhD advisor Jacob Steinhardt had me forecast AI math progress by July 2025. I predicted 30 per cent on the MATH benchmark Instead, we have IMO gold.' He credited collaborators including Sheryl Hsu and Noam Brown, and concluded by congratulating all 2025 IMO participants, noting that many OpenAI researchers are former IMO medallists themselves.
An experimental large language model (LLM) developed by OpenAI has achieved gold medal-level performance at the 2025 International Math Olympiad (IMO), setting a new benchmark in mathematical reasoning for AI systems. Announcing the milestone, OpenAI researcher Alexander Wei posted on X that the model solved five out of six problems from the latest IMO under human exam conditions. The model earned 35 out of 42 possible points, a score that would qualify for a gold medal at the real competition.
'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,' Wei explained.
The IMO is regarded as the most prestigious high school maths competition globally, known for its notoriously complex problems. Wei pointed out that such problems demand extended creative reasoning and that achieving gold-level performance represents a leap from earlier benchmarks. 'We've now progressed from GSM8K (~0.1 min for top humans) MATH benchmark (~1 min) AIME (~10 mins) IMO (~100 mins),' he said.
Submissions were graded independently by three former IMO medallists, who unanimously validated the model's solutions. According to Wei, 'the model solved P1 through P5; it did not produce a solution for P6.' He shared the model's answers publicly, noting its 'distinct style,' owing to its experimental nature.
Wei said what makes the result even more impressive is that IMO proofs are long, complex and hard to verify. "By going beyond the reinforcement learning paradigm of clear-cut, verifiable rewards we've obtained a model that can craft intricate, watertight arguments at the level of human mathematicians.'
The LLM that achieved this result will not be released publicly any time soon. Wei clarified that while OpenAI is preparing to launch GPT-5, this IMO-level model is part of a different research track. 'We don't plan to release a model with IMO gold level of capability for many months.'
OpenAI CEO Sam Altman echoed this in a follow-up post, calling the achievement 'a significant marker of how far AI has come over the past decade.' He clarified that this model is not a specialised maths system, but a general-purpose reasoning model. 'We are releasing GPT-5 soon but want to set accurate expectations: this is an experimental model that incorporates new research techniques we don't plan to release a model with IMO gold level of capability for many months,' Altman added.
Looking back, Wei also reflected on how far AI progress has exceeded expectations. 'In 2021, my PhD advisor Jacob Steinhardt had me forecast AI math progress by July 2025. I predicted 30 per cent on the MATH benchmark Instead, we have IMO gold.' He credited collaborators including Sheryl Hsu and Noam Brown, and concluded by congratulating all 2025 IMO participants, noting that many OpenAI researchers are former IMO medallists themselves. Join our WhatsApp Channel
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Time of India
23 minutes ago
- Time of India
OpenAI plans to launch GPT-5 in August: Report
OpenAI is gearing up to launch GPT‑5 in early August 2025, following a delay for additional testing. In April, Altman announced a shift in the company's AI model release schedule, revealing that intermediate models known as 'o3' and 'o4-mini' will be launched in the coming weeks, ahead of the much-anticipated GPT-5 rollout. Tired of too many ads? Remove Ads Tired of too many ads? Remove Ads Sam Altman-led OpenAI is planning to launch its much awaited advanced artificial intelligence (AI) model GPT 5 in August, according to a report by The news comes after Altman 's statement to make free copy of GPT-5 available to everybody. 'I am very interested in what it means to give everybody on Earth a free copy of GPT-5, running for them all the time," he also shared plans to roll out GPT-5 on microblogging platform X on July 19. However, he said it is not yet capable of solving the very hardest math problems, such as those in the International Math Olympiad (IMO).'We are releasing GPT-5 soon but want to set accurate expectations: this is an experimental model that incorporates new research techniques we will use in future models,' he said.'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 April, Altman announced a shift in the company's AI model release schedule, revealing that intermediate models known as 'o3' and 'o4-mini' will be launched in the coming weeks, ahead of the much-anticipated GPT-5 this month, Reuters reported that the company is also preparing to launch an AI-powered web browser that would directly compete with Google Chrome. This follows the recent release of ChatGPT Agent , a new tool that can perform tasks on users' computers autonomously.
Time of India
36 minutes ago
- Time of India
Sundar Pichai net worth: Indian-origin Google CEO becomes a billionaire as Alphabet shares near record high
Alphabet Inc.'s strong earnings report on Wednesday marked another milestone in its remarkable rally since early 2023—a surge that has added over $1 trillion to its market value and delivered a 120% return to investors. The rally has also propelled CEO Sundar Pichai into billionaire status. With Alphabet shares nearing a record high, the 53-year-old Pichai's net worth now stands at $1.1 billion, according to Bloomberg. Alphabet Inc. said demand for artificial intelligence products boosted quarterly sales, and now requires an extreme increase in capital spending — heightening pressure on the company to justify the cost of keeping up in the AI race. Explore courses from Top Institutes in Please select course: Select a Course Category PGDM Leadership Healthcare others Finance Data Science Operations Management Management Project Management Others MCA CXO MBA Degree Data Analytics healthcare Technology Design Thinking Product Management Public Policy Digital Marketing Data Science Artificial Intelligence Cybersecurity Skills you'll gain: Financial Analysis & Decision Making Quantitative & Analytical Skills Organizational Management & Leadership Innovation & Entrepreneurship Duration: 24 Months IMI Delhi Post Graduate Diploma in Management (Online) Starts on Sep 1, 2024 Get Details Google's parent company said 2025 capital expenditures will be $85 billion, or $10 billion greater than an earlier forecast. Although Alphabet beat expectations for second-quarter revenue and profit, its stock initially sank in after-hours trading, then rebounded after Chief Executive Officer Sundar Pichai explained that the investments are necessary in order to keep up with customer needs. 'Our AI infrastructure investments are crucial to meeting the growth in demand from cloud customers,' he said on a call Wednesday following the report. by Taboola by Taboola Sponsored Links Sponsored Links Promoted Links Promoted Links You May Like Bulk buying rush at Mannings! 40+ men hair changed KAMINOWA Learn More Undo As Microsoft Corp., startup OpenAI, Meta Platforms Inc. and others continue to pour money into AI, Alphabet has little choice but to follow suit, analysts said. The race is particularly urgent for Google: competitors are building chatbots that may eventually appeal to consumers more than its flagship search product. 'Google's hand is forced by OpenAI to spend tremendously on AI's infrastructure and applications,' said Nikhil Lai, an analyst at Forrester. Alphabet shares rose as much as 4.1% after markets opened in New York on Thursday, their biggest intraday gain in two months.
Time of India
40 minutes ago
- Time of India
OpenAI prepares to launch GPT-5 in August, The Verge reports
Live Events (You can now subscribe to our (You can now subscribe to our Economic Times WhatsApp channel Artificial intelligence pioneer OpenAI plans to launch its GPT-5 model as early as August, The Verge reported on Thursday, citing sources familiar with the new model, which was expected to launch this summer, will be positioned as an AI system that incorporates distinct models and can perform different functions as opposed to just a single AI did not immediately respond to a Reuters request for Microsoft-backed startup's GPT-5 will incorporate its o3 model along with other technologies, CEO Sam Altman had said in February, in a bid to simplify its startup ultimately aims to merge the o-series and GPT-series models as it looks to create AI systems that can utilize all available tools and handle a variety of tasks."While GPT-5 looks likely to debut in early August, OpenAI's planned release dates often shift to respond to development challenges, server capacity issues, or even rival AI model announcements and leaks," according to the report.
