I've been testing AI models for years now. They’re helpful. They’re tactical. They’re recently becoming strategic. But they haven’t been resonant—models with perspective so spot on I just can’t get it out of my head. o3 Pro is the first model to pass that test.
Yes, I thought o3 was a big deal. I put a lot of work into how you prompt o3. Hard to believe that was just 48 days ago. FML. At the time, o3 struck me as smart, slightly cold, and very strong on technical intelligence. Over the ~6 weeks since, it’s become an inseparable sparring partner at work.
Well, big brother just arrived. o3 Pro is a different beast altogether. The model takes ~10x as long to respond (we’re talking go get a sandwich times here). Yep, that makes it sensitive to prompts. I’ll get into that down below.
For now, I’ll just get the elephant out of the room: this is unquestionably the best model in the world right now. I get asked it a lot, and a lot of the time now the answer is “well these three are very close” and then I usually mention an OpenAI model, a Google model, an Anthropic model, and a DeepSeek model is close behind. Maybe Grok 3.
Not today. Not for a little while anyway. o3 Pro is the first model that gives me perspective I can take without filtering straight to a founder or C-suite leader. It’s sharply strategic enough (with proper prompting) to not need further polish to start a conversation about a meaningful decision. You’ll see below—we do a roadmap comparison. We also do a coding challenge, and for good measure we do a tough web research task (yes it’s about that paper). o3 Pro vs. o3, every time. o3 Pro wins easily. Every time.
I’m not saying this is a perfect model. I’ll call out some caveats toward the end of the piece. But this is absolutely a model that will give people who choose to use it well super powers. For now, it’s available on Pro and Teams accounts, but since they cut o3 pricing by 80% today, and since they’re launching o3 Pro in API for 87% less than o1 Pro, I’d expect o3 Pro to serve down (in limited quantities) to lower plans soon. Here’s what you need to know to make the most of it…
PS. Yes, you’re gonna get my full prompt plus the full responses of both o3 and o3 Pro across all three of the tests I did, plus a handy critique from Opus 4 as well, just for fun!
Listen to this episode with a 7-day free trial
Subscribe to Nate’s Substack to listen to this post and get 7 days of free access to the full post archives.