o3-mini vs DeepSeek

The AI world has been crazy lately…

First, it was Stargate, then Deepseek, and now the spotlight’s on o3 - which OpenAI released only a few days ago. Curious about this new model, I just dropped a video that puts these two head-to-head.

o3 vs. Deepseek

OpenAI’s o3-mini is their newest, most cost-efficient reasoning model, and it’s already making big claims. On the other side, we’ve got DeepSeek, which has been gaining traction for its coding and creative capabilities. But how do they really stack up?

I put them through a series of challenges:

• Advanced Reasoning: Summarizing today’s news.

• Coding: Generating a Flappy Birds game in HTML.

• Humanization: Writing an article that doesn’t sound like it was churned out by a robot.

• Website Creation: Building a modern HTML page.

The Challenges

1. Advanced Reasoning

o3-mini took the lead here. It not only delivered faster responses but also provided references (top marks). DeepSeek, while quick, didn’t categorize its thoughts as neatly - oh and as the ‘server was buys’ the info was from Octo 2023.

2. Coding (Flappy Birds Game)

This is where things got interesting. DeepSeek’s Flappy Birds game was way more functional and user-friendly. o3-mini’s attempt? Let’s just say the bird didn’t quite know where the edges of the screen were. 🐦

3. Humanization (Writing an Article)

After a few iterations, I got o3 down to 70%. However, I couldn’t get DeepSeek to work… Despite a few attempts via VPN the servers were busy, apparently. So instead, I tried with the summaries; DeepSeek got an impressive 13% but o3 got 0%! Very suprised.

4. Website Creation

o3-mini crushed this one. The HTML page it generated was sleek, modern, and ready to use. DeepSeek’s attempt felt more like a Bootstrap template—functional but not as polished.

The Verdict

So, who wins? It depends on what you’re looking for:

• o3-mini excels in reasoning, up-to-date information, and creating polished outputs like websites.

• DeepSeek shines in coding tasks (when it works) and has a more human-like flow of thought.

But here’s the kicker: DeepSeek’s server issues were a major headache. If you’re outside China, good luck getting consistent results. o3-mini, on the other hand, was reliable and fast across the board.

Want to See It in Action?

If you’re as curious as I was, you’ve got to check out the video - even if it’s just to check out Flappy Birds lol.

Luke