- OpenAI announced the upcoming launch of o3 and o3-mini AI models.
- The new model is an enhanced “inference” artificial intelligence model based on the o1 and o1-mini models released this year.
- Both models easily outperform existing AI models and will be launched in the coming months.
The last day of the event 12 days of OpenAI, bring back Open artificial intelligence CEO Sam Altman showed off a new set of artificial intelligence models that will be launched in the new year. The o3 and o3-mini models are enhanced versions of the relatively new o1 and o1-mini models. They have a natural ability to think and reason out answers before speaking. The mini version is smaller and more designed to perform a limited set of specific tasks, but the approach is the same.
Open artificial intelligence Calling it a big step toward artificial general intelligence (AGI) is a pretty bold claim, and in some ways it’s a modest improvement on an already powerful model. You may have noticed that there is a missing number between the current o1 and the upcoming o3 models. Altman said this is because OpenAI wants to avoid any confusion with British telecoms company O2.
So, what’s so special about o3? Unlike regular AI models that give answers quickly, o3 takes some time to reason about a question. This “private thought chain” allows the model to fact-check before responding, which helps it avoid some classic artificial intelligence pitfalls, such as confidently giving the wrong answer. This extra thinking time may make the o3 slower, even just a little, but the payoff is greater accuracy, especially in areas like math, science, and coding.
One of the great things about the new model is that you can manually adjust the extra think time. If you’re in a hurry, you can set it to “Low Compute” for a quick response. But if you want top-notch reasoning, turn it to “High Calculation” and give it more time to mull it over. In testing, the o3 easily outperformed its predecessor.
It’s not quite AGI; o3 cannot replace humans in every aspect. It also falls short of OpenAI’s definition of AGI, which describes models that surpass humans in the most economically valuable projects. Still, if OpenAI can pull this off, it’s comparable to Microsoft Because this will end OpenAI’s obligation to provide Microsoft with exclusive access to the most advanced artificial intelligence models.
New year, new models
Currently, o3 and its mini version are not available to everyone. OpenAI is giving security researchers a sneak peek through Copilot Labs, and the rest of us can expect the o3-mini model to be released in late January, with the full o3 model to be released shortly after. This is a careful, careful deployment, which makes sense given the capabilities and complexity we’re talking about here.
Nonetheless, o3 gives us a glimpse of where things are headed: AI can not only produce content, but actually think about problems. Whether or not it gets us toward general artificial intelligence, it’s clear that smarter, inference-driven artificial intelligence is the next frontier. Now, we’ll just have to wait and see if o3 lives up to the hype, or if OpenAI’s final gift is just a lump of coal in disguise.