In partnership with

World of AI | Edition # 43

OpenAI’s 03 Pro: Brilliant Strategist or Overthinking Giant?

OpenAI’s latest model, 03 Pro, has arrived—and it’s stirring up as much curiosity as confusion. Billed as the most powerful language model OpenAI has ever released, 03 Pro’s rollout is anything but conventional: it quietly launched alongside a significant price drop to its predecessor (03 Vanilla), operates substantially slower in real-time interactions, and displays unmatched strategic depth—if you’re willing to wait.

A Model Unlike Any Other

From a technical standpoint, 03 Pro pushes boundaries. Expert evaluators consistently rank it above the standard 03 model in science, writing, programming, and data analysis. It outperforms on benchmarks like AME 2024 and GPQA Diamond by 3%, and it achieves a remarkable 2748 ELO on Codeforces—comparable to the 159th best human competitive programmer in the world.

This performance edge, particularly in coding, suggests a model that has not only improved general capabilities but also fine-tuned its understanding of complex logic and reasoning problems. OpenAI applies a rigorous “four out of four” reliability benchmark on key tasks—demanding consistent precision across multiple attempts—and 03 Pro continues to meet this high standard.

But benchmark scores don’t tell the full story.

03 Pro is being described as a “slow thinker.” Industry users have observed it taking 10 to 25 minutes to respond to simple prompts. For example, a simple “Hi” reportedly triggered four minutes of processing. In one extreme case, it took 13 minutes to count the words in its own seven-word sentence. The explanation? Unclear—OpenAI hasn’t disclosed much about its inference pipeline or internal reasoning process.

Why Is It So Slow?

Speed issues haven’t gone unnoticed. Users like McKay Wriggley and Matt Schumer have expressed frustration over multi-minute response times. While some attribute this to deep reasoning processes, others are skeptical. “What could it possibly be thinking about for 13 minutes?” asked one reviewer.

Despite the latency, those who give 03 Pro the time it needs are often rewarded. When loaded with complex internal data, such as a company’s historical planning documents, one founder said the model generated “plausible, specific, and rooted strategic plans” that actually changed how we think about our future. That kind of result is hard to capture in a benchmark—but undeniably valuable.

The model’s depth is especially evident in specialized use cases. In one example, a researcher asked 03 Pro to critique the human immune system and propose a theoretical improvement. The model delivered a comprehensive, thoughtful breakdown of biological limitations and proposed novel solutions—far beyond the utility of a standard chatbot.

Use Cases: Power and Precision—If You Wait

03 Pro comes with full tool support out of the box: code execution, web browsing, image input, file analysis, and memory integration. It excels in strategy, science, and system-level thinking. Some notable examples include:

• Strategic Planning: Raindrop’s leadership team used it to analyze years of internal planning data and received a detailed execution plan that realigned their business direction.

• Medical Research: A doctor used 03 Pro to design a hypothetical "Immune System 2.0," and found it significantly more thoughtful than earlier versions.

• Word Puzzle Logic: It successfully solved a complex word ladder puzzle that had stumped earlier models and bested existing online solutions.

• Real-World Simulation: Flavio Adamo tested 03 Pro on a rotating ball-and-collision physics demo. It was the first model to handle realistic collisions with near-perfect accuracy.

• Jailbreaking: Despite robust refusal safeguards, some users have already jailbroken 03 Pro—demonstrating both its potential flexibility and its security vulnerabilities.

Not Without Its Failures

Still, 03 Pro is not infallible. When challenged to build a Rubik’s Cube simulation—a task previously completed by Gemini 2.5 Pro in 1,200 lines of code—03 Pro faltered with just 328 lines. After a manual fix to a coding error, the simulation ran, but the cube was rendered incorrectly and failed to rotate properly. A clever attempt, but ultimately a miss.

Even seemingly simple prompts, like counting words or generating short responses, have led to comically long delays, with users watching tokens trickle in at a glacial pace. This has raised critical questions about the model’s suitability for everyday tasks that demand quick turnarounds.

Cost and Competitiveness

Performance isn’t cheap. While 03 Pro outshines most competitors like Claude Opus 4 and Gemini 2.5 Pro in quality, it does so at a higher cost—sometimes up to $10 per task, depending on the setup. Claude Sonnet 4 and Gemini are considerably more affordable and still hold their own in many benchmarks.

There’s also the hidden cost of time. Waiting 10–20 minutes for a response—especially in interactive workflows—can be a major productivity drain unless the payoff is extraordinary. For users or organizations needing high-volume, fast-turnaround outputs, 03 Pro might not be the ideal choice.

Final Verdict: A Different Kind of Intelligence

03 Pro doesn’t shine in quick demos or surface-level tasks. It’s slow, expensive, and prone to overthinking simple prompts. But under the right conditions—when provided with complex input and ample time—it displays strategic intelligence at a level few, if any, language models can match. It’s not just an assistant. It’s closer to a thinking partner.

This shift—from task executor to strategic collaborator—marks a new chapter in AI evolution. For researchers, executives, and long-term planners, 03 Pro may prove to be an essential tool. For casual users or those in need of immediate answers, the latency and cost may be too steep.

Bottom line: 03 Pro isn’t built for speed. It’s built for depth. The magic isn’t in what you ask—it’s in how long you’re willing to let it think.

Optimize global IT operations with our World at Work Guide

Explore this ready-to-go guide to support your IT operations in 130+ countries. Discover how:

• Standardizing global IT operations enhances efficiency and reduces overhead

• Ensuring compliance with local IT legislation to safeguard your operations

• Integrating Deel IT with EOR, global payroll, and contractor management optimizes your tech stack

Leverage Deel IT to manage your global operations with ease.

Download free guide