OpenAI's new reasoning model o1-pro is powerful but pricey

OpenAI has launched a new, more powerful version of its o1 reasoning model, called o1-pro, available through the Responses API. This model offers improved responses by utilizing significantly more computational power than the original o1, making it more accurate than regular large language models like GPT-4. However, o1-pro is expensive, costing $150 per million tokens for input and $600 per million tokens generated, which is twice the price of GPT-4.5 and 10 times more than o1. It features a 200,000-token context window, a 100,000-token output cap, and supports image inputs and structured outputs. Currently, it's accessible only to select developers who have spent at least $5 on OpenAI's API services.

Despite its high cost, OpenAI believes developers will pay for the enhanced capabilities of o1-pro. The model is primarily aimed at AI agents, applications that perform tasks autonomously. An earlier version was available to ChatGPT Pro subscribers but received mixed reviews due to its limitations in solving complex puzzles and illusions. OpenAI has also developed a more advanced reasoning model, o3, which has not yet been released.

Key takeaways:

OpenAI has launched a new reasoning model called o1-pro, which is more powerful and expensive than its predecessor, o1.
The o1-pro model is available through the new Responses API and is targeted at AI agents, with pricing set at $150 per million tokens for input and $600 per million tokens for output.
Despite its high cost, o1-pro offers better responses due to increased computational power, but its other specifications remain similar to the o1 model.
An earlier version of o1-pro was tested with ChatGPT Pro subscribers, showing only slight improvements over o1 in benchmark tests, and a more advanced model, o3, is in development but not yet released.

OpenAI's new reasoning model o1-pro is powerful but pricey - SiliconANGLE

Key takeaways:

Comments (0)

Newsletter