The AI model underlying OpenAI’s self-governing web agent, Operator, has been upgraded.
Operator, an AI agent from OpenAI that can browse the web on its own and use specific software in a virtual machine hosted in the cloud to respond to user requests, is getting an update to its AI model.
One of the newest in OpenAI’s o family of “reasoning” models, o3, will soon be used by Operator. A while back Operator depended on a proprietary version of GPT-4o.
O3, a state-of-the-art addition to OpenAI’s line of “reasoning” models, will serve as the foundation for the new version.
O3 is a significantly more sophisticated model by many standards, especially when it comes to math and reasoning problems. This o3 model’s superior performance in several benchmarks, led to the decision to upgrade.
In a blog post, OpenAI stated, “We are replacing the existing GPT‑4o-based model for Operator with a version based on OpenAI o3.” “Operator’s API version will continue to be based on 4o,” even though the API version of Operator will still use the 4o model.
The enhanced Operator is one of many cutting-edge AI agents from different tech firms. These agents can carry out duties with little to no oversight.
In recent months, AI companies have released a number of agentic tools, including Operator. Businesses are rushing to create extremely complex agents that can consistently complete tasks largely unsupervised.
Google offers a “computer use” agent through its Gemini API that can similarly explore the web and execute activities on behalf of users, as well as a more consumer-focused service called Mariner. Additionally, Anthropic’s models can navigate websites and open files, among other computer functions.
In order to “teach the model [OpenAI’s] decision boundaries on confirmations and refusals,” the new Operator model, known as o3 Operator, was “fine-tuned with additional safety data for computer use,” according to OpenAI.
Additionally, Anthropic offers models that can navigate web pages and open files, among other computer functions.
A technical report from OpenAI details how well o3 Operators performed on particular safety assessments. According to the technical paper, o3 Operator is less likely than the GPT-4o Operator model to decline to engage in “illicit” activities and look for private information. It is also less vulnerable to prompt injection, a type of AI attack.
In a blog post, OpenAI stated that “o3 Operator employs a similar multiple layers strategy for safety that was adopted for the 4o version of Operator.” “O3 Operator does not have native access to a coding environment or terminal, even though it inherits o3’s coding capabilities.”