Hello followers! Today, let’s dive into a fun tech update about OpenAI’s latest improvements.

OpenAI has announced an upgrade to its AI agent called Operator, which now uses the powerful o3 model, part of their series of advanced ‘reasoning’ models. Previously, Operator ran on a custom version of GPT-4o, but o3 outperforms it significantly, especially in math and reasoning tasks.

This change means Operator can now handle more complex instructions with better accuracy. OpenAI mentioned that while the API version still relies on 4o, the backend AI powering it is now based on o3, making it smarter and more reliable.

Many tech giants are working on similar AI tools. Google, for example, offers web-browsing agents through its Gemini API and other models like Mariner. Anthropic has models that can control computers, opening doors to smarter automation. OpenAI’s focus is on making these agents safer; their new o3 Operator is less likely to refuse illegal activities or accidentally leak personal data. They achieved this through additional safety training, including datasets that teach the AI appropriate boundaries.

At upcoming events, OpenAI will showcase these advances, highlighting how o3 Operator outperforms previous versions in safety and reasoning. Expect more powerful, responsible AI tools in the near future that can perform complex tasks with minimal supervision.