OpenAI is updating the AI mannequin powering Operator, its AI agent that may autonomously browse the net and use sure software program inside a cloud-hosted digital machine to satisfy customers’ requests.
Soon, Operator will use a mannequin primarily based on o3, one of many newest in OpenAI’s o sequence of “reasoning” fashions. Previously, Operator relied on a customized model of GPT-4o.
By many benchmarks, o3 is a much more superior mannequin, notably on duties involving math and reasoning.
“We are changing the prevailing GPT‑4o-based mannequin for Operator with a model primarily based on OpenAI o3,” OpenAI wrote in a weblog submit. “The API model [of Operator] will stay primarily based on 4o.”
Operator is one amongst many agentic instruments launched by AI firms in current months. Companies are racing to make extremely refined brokers that may reliably perform chores roughly with out supervision.
Google presents a “pc use” agent by way of its Gemini API that may equally browse the net and take actions on behalf of customers, in addition to a extra consumer-focused providing known as Mariner. Anthropic’s fashions are additionally in a position to carry out pc duties, together with opening recordsdata and navigating net pages.
According to OpenAI, the brand new Operator mannequin, known as o3 Operator, was “fine-tuned with further security information for pc use,” together with datasets designed to “train the mannequin [OpenAI’s] resolution boundaries on confirmations and refusals.”
OpenAI has launched a technical report displaying o3 Operator’s efficiency on particular security evaluations. Compared to the GPT-4o Operator mannequin, o3 Operator is much less more likely to refuse to carry out “illicit” actions and seek for delicate private information, and fewer inclined to a type of AI assault often called immediate injection, per the technical report.
“o3 Operator makes use of the identical multi-layered strategy to security that we used for the 4o model of Operator,” OpenAI wrote in its weblog submit. “Although o3 Operator inherits o3’s coding capabilities, it doesn’t have native entry to a coding atmosphere or terminal.”
