OpenAI has launched o1-Pro, its most powerful model yet. Currently, the model is only available to select developers on tiers 1-5.
This model supports vision, function calling, and Structured Outputs, and works with the Responses and Batch APIs.
And it provides better responses compared to the o1. It does this by thinking harder by using more computing.
The pricing for this model, on the other hand, is wild. $150 for 1M input tokens and $600 for 1M output tokens. With a 50% off on Batch API pricing. This makes the o1-pro the most expensive model OpenAI has to offer. And with good reason.
The only downside is that the model is the slowest, too. Of course, due to the heavy computational power behind it.
A comparison from OpenAI’s website is below.
The catch?
OpenAI pro subscribers are reported saying that the o1-pro is prone to mistakes. People were disappointed by the $200 model.
For example, giving o1-pro to draw a unicorn and failing. This raises many questions. However, a pertinent one is: Do the capabilities of AI blind us to their massive shortcomings?