O3’s Most Surprising Use Cases to Transform Your Work

Just days after OpenAI released their new O3 and O4 Mini models, I've been amazed at how quickly professionals across industries are discovering groundbreaking applications. As someone who closely follows AI developments, I'm seeing capabilities that go far beyond incremental improvements—these models represent a genuine leap forward that business leaders need to understand.

The Intelligence Leap Is Real

The industry reaction has been strong, and with good reason. O3 has achieved the highest IQ score of any AI model on Earth—136 on the Mensa IQ test, significantly exceeding Gemini 2.5 Pro's previous high of 128. To put this in context, this places O3 in the "gifted" range of human intelligence, higher than roughly 98% of the population. For reference, Einstein was estimated to have an IQ of 160.

What makes this relevant for businesses isn't just the impressive benchmark numbers, but how this intelligence translates to practical problem-solving. As AI expert Daria Enutz noted after early access to the model: "When I throw challenging clinical or medical questions at O3, its responses sound like they're coming directly from top subspecialist physicians. Precise, thorough, confidently evidence-based, and remarkably professional."

Tool Usage Within Reasoning: The Game-Changer

The most transformative capability I've seen—and what I believe will revolutionize how professionals work—is O3's ability to use tools within its reasoning chain.

When Amjad Msad, CEO of Replet, demonstrated this feature, I immediately recognized its significance for business users. The model was asked to calculate the average compound growth rate of Apple's stock. Instead of providing a generic response, it wrote Python code within its thinking process, executed that code to perform the calculation, and delivered the precise answer.

This means the model can now:

• Identify when computational tools are needed
• Write the appropriate code on the fly
• Execute that code within its reasoning process
• Interpret the results and continue the analysis

For business professionals, this eliminates the need to switch between different tools or rely on technical specialists for many tasks. Your AI assistant can now handle multi-step analytical processes end-to-end. It can use tools like web search, code interpreter, reminders, and memory in a loop so you can have it code complex features, or even answer tricky research queries over long documents.

Solving Previously Impossible Challenges

What's truly fascinating is seeing O3 tackle challenges that were previously considered AI-resistant:

1.     Mini Courses: O3's Breakthrough in Personalized Learning

One of the most exciting O3 use cases I've seen is its ability to create personalized mini-courses with scheduled daily lessons. Before O3's reasoning and tools capabilities, AI models couldn't effectively sequence educational content over time or provide ongoing structured learning experiences. They could generate one-off lessons but lacked the ability to plan a coherent curriculum and deliver it systematically.

O3 changes this by combining deep domain knowledge with its reminder tool functionality, allowing it to design a comprehensive course (like machine learning fundamentals) and then automatically deliver bite-sized, sequential lessons at scheduled times. This creates a personalized learning experience that builds progressively over days or weeks, transforming the model from a passive information provider into an active, ongoing education partner.

2. Location Intelligence

O3 has effectively "solved" the challenge of determining geographic locations from limited visual information. When given a random Street View image, it can identify not just the country, but often the specific region or city, with no prior training on this specific task.

For businesses, this has immediate applications in market research, competitive intelligence, and geographic analysis. Imagine being able to quickly analyze property listings, competitor locations, or regional market trends through visual data.

It can even figure out a place from some of its attributes, such as a specific dish in a restaurant.

3. Mathematical Problem-Solving

Scott Swingingle reported that O4 Mini High solved the latest Project Euler problem (a notoriously difficult mathematical challenge) in just 2 minutes and 55 seconds—faster than any human solver. Only 15 humans were able to solve it in under 30 minutes, with the fastest human taking 5 minutes and 15 seconds.

For financial analysts, data scientists, and strategic planners, this means complex optimization problems that once required specialized expertise can now be tackled more efficiently.

4. Visual Reasoning and Coding

The models excel at tasks requiring both visual understanding and coding implementation. Flavio Adamo's hexagon and balls test, which evaluates a model's ability to understand physics and implement it in code, was flawlessly passed by both O3 and O4 Mini.

For developers and technical teams, this translates to more accurate code generation, better debugging assistance, and reduced development time.

The Business Value Is Clear

According to artificial analysis's independent benchmarks, O4 Mini has made particularly impressive gains in coding intelligence, achieving the number one position in the coding index with a 7% point improvement in both live codebench and sciode tests.

For business leaders, this means:

1. Higher-quality code with fewer bugs
2. Faster development cycles
3. Less reliance on specialized programming resources
4. More accurate implementation of complex business requirements

Not Without Limitations

Despite these breakthroughs, it's important to acknowledge the models' limitations:

1. Context Window Constraints: O4 Mini's context window of 200K tokens remains unchanged from O3 Mini, notably smaller than Claude's and Gemini's larger context windows. This limits the amount of information the model can process at once.
2. Visual Understanding Gaps: Some tests reveal persistent challenges with detailed image analysis. When asked to identify people and colors in a drawing, O3 still made basic identification errors.
3. Occasional Basic Errors: As Bojan Tunguz, ex-NVIDIA, demonstrated, the model sometimes still misses simple questions like counting the number of "r"s in "strawberry."

Why This Matters for Your Work

As professionals navigating the AI revolution, understanding these capabilities helps position you to leverage them effectively. The improvements in mathematical reasoning, coding, and tool integration mean you can now automate more complex analytical tasks, generate higher-quality code, and solve problems more efficiently.

The rapid pace of innovation we're seeing—with models demonstrating capabilities that seemed years away just months ago—means organizations need to continually reassess their AI strategy. The companies that thrive will be those that quickly identify and implement these new capabilities to solve real business problems.

What's Next?

• Explore our new best-selling book, Agentic Artificial Intelligence, to learn how to successfully implement agents in your organization, work, and life.
• Join the first Executive Course on Agentic AI to learn how to successfully integrate it in your organization and build stronger human-AI collaboration.
• Discover IRREPLACEABLE for practical ways to grow the human strengths that matter most in an AI-powered future.

Thanks,

-- Pascal