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
Hans India
15 minutes ago
- Hans India
Google introduces AI Skill Academy in India
New Delhi: Tech giant Google on Monday said that it has launched the Google News Initiative (GNI) AI Skills Academy, in collaboration with the Indian Institute of Mass Communication (IIMC) here. According to the company, the new initiative is aimed at equipping Indian newsrooms with the knowledge and tools they need to thrive in an AI-powered future. "Continuing our commitment to collaborate with news organisations across India and bring them Google's best-in-class technology, we're excited to announce the launch of the Google News Initiative AI Skills Academy in collaboration with the Indian Institute of Mass Communication (IIMC), Department of New Media," the tech giant said in a statement. This will be a 10-week, hybrid training series, which is designed to equip newsrooms and media educators with foundational AI understanding and practical skills. Participants will learn to leverage Google's AI tools like NotebookLM, Gemini, AI Studio and Pinpoint to streamline workflows, boost efficiency, and free up valuable time for deeper and more creative research and in-depth, diverse storytelling. Launched by Google in an academic partnership with IIMC and with training support from How India Lives, this hybrid programme will empower participants to apply AI tools across a range of relevant use-cases. The programme will offer weekly deep dives, practical exercises, dedicated mentoring, and problem-solving sessions. This programme has been curated to provide participants with support to leverage AI to perform newsroom tasks more efficiently. "We're also proud to support IIMC in training media educators and students across its campuses in six cities in India," Google stated. This collaboration is a major step towards empowering media professionals and media educators with essential AI skills. "As AI transforms journalism, this initiative will help them stay ahead. We intend to promote responsible innovation and enhance creativity in storytelling. IIMC is happy to be part of this initiative that will also help train students across our six campuses', said Nimish Rustagi, Registrar, Indian Institute of Mass Communication.
Hans India
15 minutes ago
- Hans India
Trump's AI Tool Targets Massive Federal Regulation Cuts, Sparks Legal and Ethical Debate
In a bold move to reshape the federal regulatory landscape, the Trump administration has turned to artificial intelligence to fast-track its ambitious deregulation agenda. A new report from The Washington Post reveals that a government-deployed AI system, dubbed the DOGE AI Deregulation Decision Tool, is being used to identify and eliminate a sweeping number of federal rules—potentially as many as half of the current 200,000 regulations. The tool is operated under the Department of Government Efficiency (DOGE), a body created to modernize and streamline federal operations. So far, the AI has already reviewed and recommended removal of more than 1,000 regulations at the Department of Housing and Urban Development (HUD) in just two weeks. It's also credited with drafting all recent deregulations at the Consumer Financial Protection Bureau (CFPB). Government insiders told The Post that DOGE AI was developed by a team of engineers recruited during tech billionaire Elon Musk's brief involvement with the agency. According to a presentation reviewed by the publication, DOGE AI is being promoted as a cost-saving solution that could cut bureaucratic red tape, lower compliance costs for businesses, and attract new investment by simplifying the regulatory environment. Agencies across the federal government have been given a deadline of September 1 to submit their lists of regulations to be reviewed—and potentially scrapped—using the AI tool. The Trump administration hopes this initiative will deliver visible results by the first anniversary of Trump's return to office. This AI-driven approach follows Trump's earlier executive order, issued in January, directing federal agencies to eliminate ten existing rules for every new one introduced. Departments like Transportation and Labor have already announced significant rollbacks of existing regulations as part of this push. Despite the technological enthusiasm, the move has drawn mixed reactions across federal agencies. While some departments have embraced DOGE AI's rapid processing capabilities, others are voicing caution. Critics argue that relying on AI to review intricate and legally sensitive regulations could result in oversight, errors, or even violations of administrative law. Legal experts emphasize that repealing federal rules is not a simple task. Administrative law mandates rigorous processes, including public consultations, environmental impact assessments, and legal reviews. Automating this work, they argue, could undermine the integrity of the system. Adding to the uncertainty is internal tension among federal staff. Some employees fear that increased dependence on AI could lead to flawed policy decisions. Meanwhile, ongoing staffing cuts are reportedly hampering the speed at which agencies can review or respond to AI-generated suggestions, despite pressure from the White House for faster results. Still, the administration maintains confidence in the technology. 'We're exploring all options,' said White House spokesperson Harrison Fields, adding that while nothing is finalized, the DOGE team deserves credit for introducing fresh ideas into government operations. As the deadline approaches, the ultimate impact of DOGE AI remains unclear. But what's certain is that this experiment in algorithmic governance is already reshaping conversations about the future of policymaking in Washington. TAGS: Trump deregulation AI tool, DOGE AI federal rules, US regulation cuts 2025, Technology, Tech News
Economic Times
an hour ago
- Economic Times
TCS to layoff 12,000 employees: What we know so far about the mass cull
Tata Consultancy Services, the largest software exporter in India, has announced layoffs that will affect around 12,000 employees, or 2% of its global workforce. The announcement follows suspense on salary hikes after a tepid June quarter, anxiety among staff over a new benching policy, and employee bodies decrying delays in lateral hiring. Here's a detailed look at all the issues: TCS layoffs: All you need to know The massive layoffs will primarily affect employees in middle and senior roles. Employees on the bench for a long time may also face the heat. The layoffs will happen across TCS's global workforce and will not be focused on any specific geography or industry domain. The IT major will complete the job cuts over the next three quarters of FY26. Job cuts for agility, not because of AI: CEO In a statement on the job cuts, TCS said it is on a journey to become a 'future-ready organisation', which calls for realigning the workforce model, among other things."Towards this, a number of reskilling and redeployment initiatives have been underway. As part of this journey, we will also be releasing associates from the organisation whose deployment may not be feasible," the company said. Talking to Moneycontrol, TCS chief executive K Krithivasan emphasised that the retrenchment is not a case of AI taking jobs, but rather a skill mismatch issue. The company has been training employees in AI skills, but senior-level staff may not be able to use entry-level skills, causing issues with their deployment, the TCS CEO stated. What TCS is offering affected employees The laid-off employees will receive payments for their notice periods, along with a severance package, TCS said. The company said it is also looking to extend insurance benefits and other outplacement opportunities for the impacted employees. In its statement, the company said it will 'provide appropriate benefits, outplacement, counselling, and support' to employees 'as they transition to new opportunities.'CEO Krithivasan told Moneycontrol that the company will not rush the process. It will first talk to the employees identified for layoffs, and try to deploy them. If unable to do so, the fired staff will be given benefits as per the HR policy, he said.
