OpenAI o3 Mini

OpenAI o3 Mini

OpenAI Just Dropped a New Budget-Friendly AI model. OpenAIs new o3-mini is like a smarter, faster cousin of its earlier AI models like the o1 or o1 mini. Its built for complex STEM (Science, Technology, Engineering & Maths) tasks. You can ask this model to do your maths homework or a summer science project and the possibilities are endless. Open AI has claimed that this model is cheaper and quicker than its earlier model and is also said to be faster than the trending model of the current DeepSeek. Let's take a technical look at the model and the price comparison with its direct rival DeepSeeks R1 model.

Whats Interesting About o3-Mini?

As per OpenAI, the model facts check itself with access to the latest information. In simple words, it double-checks the results it gives to the user.

The model in medium reasoning effort mode is equal to the o1's performance in maths, coding, and science with a faster response time! There is a new thing with the model which is called reasoning effort which in human terms can be called the brain power of the model which has 3 modes: low, medium, and high which can be adjusted as per the user's usage. It leverages this technique to balance the speed and accuracy of the given task.

Even with the double, it is 24% faster than the OpenAI's older o1-mini model and costs 63% cheaper, which is a great thing.

Let's take a price comparison of the model

Model	Input Cost	Output Cost
o3-mini	$0.55	$4.40
o1-mini	$1.50	$12.00
o1	$3.00	$15.00
DeepSeek R1	$0.14	$2.19

Via this, we can easily say that the o3 mini is far cheaper than the earlier models released by OpenAI but is not that cheap when compared to the Deepseek R1 model.

Note: The R1 model is China and we all know how much China is hungry for the data of the people which is not safe. You can run the R1 model locally on your system. You can check the blog here to learn how to do that.

Is there a competition between R1 and o3?

Yes, there are trades in both the models, o3 beats R1 in complex tasks like coding but costs way more than the R1 but it lags a little bit when doing some PhD level stuff which normal users won't be doing but still that's a loss for some users.

The R1 model is cheaper and faster than the o3 to some extent but the issue is that it is China-based and once they acquire your data you don't know how they can use it.

In the end, the o3 model is great but comes with its caviars while the R1 model is awesome and cheap but is China-based and cannot be trusted that much.