It is released

Openai has Firing O3-PRO, AI model that the company claims to be the most capable yet.

O3-PRO is a copy of Openai’s O3Thinking model that was launched earlier this year. Unlike traditional artificial intelligence models, thinking models work through step -by -step problems, enabling them to perform more reliable in fields such as physics, mathematics and coding.

The O3-PRO is available for Chatgpt Pro and team users that start on Tuesday, to replace O1-PRO model. Enterprise and EDU users will get access to the following week, says Openai. The O3-PRO also lives in the API developer Openai from the afternoon of this day.

The price of O3-PRO is $ 20 per million input codes and $ 80 per million API output symbols. Distinctive icons of input are the distinctive symbols that fed the model, while the distinctive symbols are the symbols created by the model based on the distinctive icons of input.

A million inputs equivalent to about 750,000 words, slightly longer than “war and peace”.

“In expert evaluations, the auditors are constantly preferred to O3-PRO over O3 in each category tested, especially in major fields such as science, education, programming, business and writing,” he writes Openai in ChangeLog. “The auditors have also constantly rated O3-PRO for clarity, comprehensiveness, and follow-up of instructions and accuracy.”

O3-PRO has access to tools, according to Openai, which allows them to search in the web, analyze files, the reason for visual inputs, use Python, customize their memory responses, and more. As a sign, the model usually takes longer than the O1-PRO to complete, according to Openai.

O3-PRO has other restrictions. Temporary chat is disabled with the model in ChatGPT at the present time while Openai “is a technical problem”. The O3-PRO images cannot generate. And Canvas, the Acts Opening Open, is not supported by O3-PRO.

On the positive side, O3-PRO achieves impressive degrees in common artificial intelligence standards, according to Openai’s internal test. At AIME 2024, which holds mathematics skills in the model, the O3-PRO grades are better than the higher performance artificial intelligence model from Google, Gemini 2.5 Pro. O3-PRO also outperforms the recently released Anthrobour Claude 4 OPUS On GPQA Diamond, a doctoral scientific knowledge test.