The Director General of OpenAi Sam Altman (Sam Altman) exactly two weeks ago promised that the large language model of the new generation O3-Mini with the ability to reasoned would be released “in a couple of weeks”. And he didn’t deceive-today Openai launched O3-Mini in ChatGPT, as well as in API services. The most interesting thing is that the new AI model has become available even to free ChatGPT users, albeit with restrictions.
Image source: OpenAI
The premiere of Openai originally announced within the framework of the 12-day pre-science marathon, the O3-Mini model is designed to compare with O1 in terms of performance in tasks related to mathematics, writing software code and scientific disciplines, while responding faster. Openai claims that O3-Mini works 24 % faster than O1-Mini, and at the same time gives more accurate answers. Like O1-Mini, the new model will demonstrate the course of its thoughts when solving the problem, and not just provide a ready-made answer.
In December, Openai introduced several early tests demonstrating the superiority of O3 over O1. Now the developers argue that the O3-Mini version will surpass O1 in a number of tasks related to coding and reasoning, with lower costs and delays. Developers will be able to use O3-Mini through the Openai API services, including the Chat Complets API, Assistants API and Batch API.
The O3-Mini-High model will also be available to paid users, which, according to Openai, will be the “best option for writing software code in Chatgpt” and will offer answers with a higher level of intelligence, albeit with a slight delay. In addition, O3-Mini will support the search on the Internet, allowing you to find answers with links to web sources.
This is also the first time when free ChatGPT users will be able to test the Openai models with the ability to reason. Probably, for this it is worth thanks to the Chinese startup Deepseek, which stirred the world of AI. Although the influence of Microsoft, which had previously opened access to O1 for all Copilot users, cannot be ruled out.
Users will be able to test the O3-Mini in ChatGPT for free by choosing Reason function on the chat panel. The restrictions will be the same as for the GPT-4O. For paid users, the limits will be higher: ChatGPT Plus and Teams tariff subscribers will be able to send up to 150 messages per day. And users of the ChatGPT Pro tariff for $ 200 per month will receive unlimited access to O3-Mini.