OpenAI Launches O1 AI Model That ‘Thinks’ Before It Answers

OpenAI has introduced its latest AI model, OpenAI O1, part of its secretive “Project Strawberry.” The O1 model is the first in a new series of reasoning AI systems designed to handle complex tasks in coding, mathematics, and science. This launch represents a significant leap forward in AI technology, aiming to mimic human-like problem-solving capabilities.

How Does the O1 Model Work?

OpenAI’s O1 model stands out for its ability to “think” through queries before answering. Much like a human solving a difficult problem, O1 approaches tasks from various perspectives, checks its work, and learns from its mistakes. This is a key improvement over earlier models, which often provided quick but less thoughtful responses.

In tests, the O1 model performed at the level of PhD students in subjects like physics, chemistry, and biology. For example, it solved 83% of problems in a tough maths contest, compared to just 13% by previous versions. In coding, O1 ranked higher than 89% of participants, making it a game-changer for software developers.

O1-Mini: A Cheaper and Faster Version

To complement O1, OpenAI has also released O1-Mini, designed for developers who need a faster, more affordable model. The O1-Mini is 80% cheaper than the O1-preview version, making it a cost-effective option while retaining strong reasoning and coding capabilities.

Enhanced Safety Features

OpenAI has introduced a new training method to make O1 safer to use. The model is better at following safety guidelines, even when users attempt to “jailbreak” it by tricking it into breaking rules. In a recent safety test, the earlier version of the AI scored just 22 out of 100, while the new O1 model scored 84—a massive improvement.

OpenAI has collaborated closely with the UK and US governments, running extensive safety tests and gathering input from experts to improve the model’s safety measures.

Impact on Jobs and Research

O1’s ability to solve complex tasks could impact jobs in areas like software development, data analysis, and coding. With O1 able to perform at the level of highly skilled professionals, there may be reduced demand for routine tasks, pushing workers to focus on higher-order thinking skills like creativity and critical analysis.

For researchers, O1 is a valuable tool, particularly in fields like healthcare and science. Its ability to analyse large datasets and generate complex formulas could accelerate breakthroughs.

How to Access OpenAI O1

Starting today, OpenAI O1 can be accessed by ChatGPT Plus and Team users, with a limit of 30 messages per week for O1-preview and 50 for O1-mini. The company plans to increase these limits soon, and enterprise users will gain access next week.