OpenAI upgrades the AI model powering its Operator agent

OpenAI is updating its AI agent, Operator, to use a new model called o3, which is part of OpenAI's latest series of "reasoning" models. This update replaces the previous GPT-4o-based model and is expected to enhance Operator's performance, particularly in tasks involving math and reasoning. The o3 Operator model has been fine-tuned with additional safety data to improve its decision-making capabilities, making it less likely to engage in illicit activities or fall victim to prompt injection attacks. Despite inheriting o3's coding capabilities, the o3 Operator does not have native access to a coding environment or terminal.

This update is part of a broader trend among AI companies to develop sophisticated agents capable of autonomously performing tasks with minimal supervision. Competitors like Google and Anthropic are also releasing similar agentic tools. OpenAI has released a technical report detailing the safety evaluations of the o3 Operator, highlighting its improved performance over the previous model.

Key takeaways

OpenAI is updating its AI agent, Operator, to use a model based on o3, which is more advanced in math and reasoning tasks compared to the previous GPT-4o-based model.
The new o3 Operator model has been fine-tuned with additional safety data to improve decision-making on confirmations and refusals.
o3 Operator is less likely to engage in illicit activities or be susceptible to prompt injection attacks, according to OpenAI's technical report.
Despite its advanced capabilities, o3 Operator does not have native access to a coding environment or terminal.

OpenAI upgrades the AI model powering its Operator agent | TechCrunch

Key takeaways

Discussion (0)