(Untitled)

Artificial intelligence models developed by OpenAI and Google DeepMind have each earned gold-medal level scores in the 2025 International Math Olympiad (IMO) — a landmark global math competition for high school students. Both companies recently announced these achievements, showcasing just how rapidly AI systems are progressing in complex reasoning and problem-solving tasks.

The AI Race Enters the Math Arena

This year, both OpenAI and Google submitted what are known as “informal” AI systems to the IMO challenge. Unlike previous years, where AI models required human intervention to translate problems into code or formal logic, the latest entrants tackled math questions directly in natural language—reading the prompts, reasoning through the solutions, and generating proof-based answers just like real students. According to the companies, their systems not only surpassed last year’s AI benchmarks but also outperformed most human participants worldwide.

Disputes Over AI’s Moment in the Spotlight

Although both companies’ models performed exceptionally, the celebration soon turned into a debate. Google publicly questioned how OpenAI timed and announced its results—highlighting that OpenAI’s test wasn’t officially evaluated by IMO judges and suggesting that only officially graded results should claim gold-medal status. Google, for its part, closely coordinated with IMO organizers, announcing its achievement only after official grading and results were released. This back-and-forth highlights just how important public perception is in the AI race, especially when it comes to impressing the best AI research talent—many of whom come from competitive math backgrounds.

AI Benchmarks: More Than a Score

OpenAI explains that it had independent third-party evaluators (former IMO medalists) assess its models using IMO’s grading standards, while Google relied on the official IMO evaluators. OpenAI claims it didn’t know that Google was participating in the same informal testing format, and that they only announced after verifying with IMO officials. Regardless of the process, the results from both labs are a powerful signpost: AI is closing the gap with human competitors in complex cognitive domains beyond language and data recall.

Deep Founder Analysis

Why it matters

For startups and technology innovators, this development is an inflection point. It signals that AI is no longer limited to predictable, data-heavy tasks, but is making real headway into domains that demand abstract reasoning, creativity, and multi-step problem-solving. Startups building tools for education, research, or even technical hiring should pay close attention, as these benchmarks will likely set industry standards — and customer expectations — for what “intelligent” systems can achieve.

Risks & opportunities

The rapid pace of progress also creates both opportunities and risks. While startups could leverage AI with near-expert math or analytical skills to disrupt traditional processes (like automated tutoring, STEM assessment, or even technical interviews), there is a danger of over-reliance on AI when critical evaluation is still evolving. Companies must monitor for model inaccuracies or unresolved “black box” behaviors, echoing the way search engines changed but also complicated information access. Moreover, the competitive tension between industry giants can create winner-take-all dynamics, making it challenging for challengers to keep up without niche differentiation.

Startup idea or application

Inspired by these results, a promising startup could create an adaptive math learning platform that combines state-of-the-art AI models with real-time feedback loops. Imagine a product that not only explains solutions, but dynamically evaluates a student’s reasoning process, offers alternative strategies, and adjusts difficulty in a way that rivals or even surpasses top human tutors. Alternatively, startups might build tools that help companies objectively assess candidates’ technical abilities using AI-driven competitions or challenges similar to the IMO format, expanding fair access to global talent.

The Takeaway: A Tight Race Signals a Paradigm Shift

While OpenAI was once considered clearly ahead in the AI field, Google’s recent strides in formal recognition mean the competition is much closer than before. Both companies’ current-generation systems performed at levels comparable to only the top few percent of IMO competitors worldwide. As OpenAI eyes the release of GPT-5, the rivalry is likely to intensify—pushing AI’s boundaries ever further.

AI Google OpenAI Competitive Math Startup Insights

For more AI milestones and their impact on startups, see our related article Perplexity Charts a Bold Path in India to Challenge OpenAI.

Visit Deep Founder to learn how to start your own startup, validate your idea, and build it from scratch.

📚 Read more articles in our Deep Founder blog.