OpenAI powers up Operator with advanced o3 model for smarter web

In a remarkable development, OpenAI has announced an upgrade to the artificial intelligence (AI) model powering its Operator agent, enhancing its ability to perform tasks that require reasoning, web browsing, and software use within a cloud-hosted virtual environment.

The upgraded Operator will soon run on a model based on OpenAI’s latest o3 architecture, part of its new line of reasoning-focused models.

Previously, Operator relied on a custom version of GPT-4o. According to the Sam Altman-led AI startup, o3 outperforms its predecessor in several benchmarks, particularly those involving complex reasoning and mathematics.

“We are replacing the existing GPT-4o-based model for Operator with a version based on OpenAI o3,” the company stated in a blog post.

However, it noted that the API version of Operator will continue to use GPT-4o.

For those unversed, Operator is a part of a growing wave of AI agents developed to carry out complex tasks with minimal user supervision.

The new version of Operator, called o3 Operator, has been fine-tuned with extra safety training focused on computer usage.

According to OpenAI, this includes specially curated datasets to help the model better understand when to confirm, refuse, or take specific actions based on user input.