OpenAI is upping the stakes for artificial intelligence reasoning applications with the debut of a more powerful version of its popular o1 model.

The company has just launched o1-pro, making it available through its new developer application programming interface called Responses API.

According to OpenAI, the o1-pro model can provide “consistently better responses” than the original o1 reasoning model because it uses significantly more computational power. AI reasoning models are said to be much more accurate than regular large language models such as OpenAI’s GPT-4, because they take more time to consider how they should respond to user’s prompts.

For now, OpenAI is making o1-pro available only to select developers who have spent at least $5 on its API services, and it’s very expensive to access.

The company has priced o1-pro at $150 per million tokens, which equates to around 750,000 words fed into its model, and $600 per million tokens generated. That makes it twice as pricey as OpenAI’s most powerful regular mode, GPT-4.5, and 10 times more expensive than the o1 model. It’s also a staggering 10,000 times more expensive than its most affordable model, GPT-4o-mini.

For the extra money, the only thing it provides is more computing power, and therefore better responses. The rest of its specifications are similar to o1, with a 200,000-token context window, a 100,000 cap on output tokens, and a Sept. 30, 2023 knowledge cut-off date. It also supports image inputs, function calling, meaning it can be connected to external data sources, and structured outputs, a feature that enables developers to ensure its outputs are generated in a specific data format.

For now, o1-pro is available only through the new Responses API, which means it’s primarily targeted at so-called AI agents, which are applications designed to perform tasks autonomously on behalf of users. Developers who have built applications using OpenAI’s Chat Completions API cannot currently access o1-pro.

Despite o1-pro being considerably more expensive than o1, OpenAI seems to think that some developers will be happy to pay those higher prices.

“O1-pro in the API is a version of o1 that uses more computing to think harder and provide even better answers to the hardest problems,” a spokesperson for the company told TechCrunch. “After getting many requests from our developer community, we’re excited to bring it to the API to offer even more reliable responses.”

The company posted a bunch of screenshots on X showing the numerous demands it had from the developer community for a more powerful version of o1 with API access, but it remains to be seen if those users will be satisfied.

o1-pro now available in API @benhylak @literallyhimmmm @shl @joshRnold @samgoodwin89 @byamadaro1013 @adonis_singh @alecvxyz @StonkyOli @gabrielchua_ @UltraRareAF @yukimasakiyu @theemao @curious_vii

It uses more compute than o1 to provide consistently better responses. Available… pic.twitter.com/Iub6tp1NTi

— OpenAI Developers (@OpenAIDevs) March 19, 2025

An earlier version of o1-pro was made available for ChatGPT Pro subscribers in December, but many users complained that it was far from perfect, struggling with Sudoku puzzles and unable to perceive optical illusions, for example.

In benchmark tests results published in December, the o1-pro model delivered only slightly better results than o1 when challenged with math problems and coding tasks.

OpenAI has also developed a more advanced reasoning model called o3, but it has not yet been released.

Image: SiliconANGLE/Dreamina