o3-mini Crushes DeepSeek R1? A Python Program Attracts Nearly 4 Million Onlookers
OpenAI’s new o3-mini outshines DeepSeek R1 in physics simulations, delivering superior ball collision and realism in complex geometric challenges.
"AI Disruption" publication New Year 30% discount link.
The headlines in the AI community were dominated by DeepSeek for over a dozen days. Yesterday, OpenAI finally couldn’t sit still any longer and launched its brand-new series of inference models, o3-mini. Not only did they open up their inference models to free users for the first time, but compared to the previous o1 series, the cost has been reduced by as much as 15 times.
OpenAI also claims that this is the newest and most cost-effective model in their inference model series.
Almost immediately after the release, some users couldn’t wait to compare it with China’s large model, DeepSeek R1, which has swept through the entire large-model circle.
Results of o3-mini on Humanity's Last Exam.
A while back, the AI community became engrossed in using DeepSeek R1 and other (inference) models to tackle the following challenge:
“Write a Python script that makes a ball bounce inside a certain shape. Let the shape rotate slowly, and ensure the ball remains within the shape.”
This simulation of a bouncing ball is a classic programming challenge. Essentially, it involves a collision detection algorithm that requires the model to recognize when two objects (for example, a ball and a side of a shape) collide. A poorly written algorithm would exhibit obvious physical errors.