OpenAI’s ChatGPT Pro Gets $200 Upgrade With New o3 Operator
▼ Summary
– OpenAI upgraded its Operator autonomous web agent in ChatGPT, switching from GPT-4o to the more advanced o3 reasoning model, available as a research preview for ChatGPT Pro subscribers.
– Operator, launched in January 2025, is a semi-autonomous agent designed to perform web-based tasks like bookings and data gathering via a cloud-hosted virtual browser for safety.
– The o3 model improves accuracy, task success rates, and response clarity, with notable performance gains on benchmarks like OSWorld, WebArena, and GAIA.
– Safety measures remain, including user confirmations for sensitive actions and restrictions on high-risk platforms, with reduced prompt injection vulnerabilities.
– The upgrade benefits enterprise professionals by enhancing workflows in AI engineering, data management, and security, offering automation with built-in safeguards.
OpenAI has rolled out a significant upgrade to its ChatGPT Pro service, introducing enhanced capabilities through its new o3 reasoning model. The $200 monthly subscription now includes access to an improved version of Operator, the company’s autonomous web browsing tool, marking a strategic move in the competitive AI landscape. This update positions OpenAI favorably against rivals like Google, whose premium AI bundle typically costs nearly $250 per month.
Operator first launched in early 2025 as OpenAI’s foray into semi-autonomous agents, designed to perform web-based tasks like booking reservations or compiling shopping lists. Unlike traditional browser extensions, it operates in a secure, cloud-hosted environment, ensuring user privacy while executing actions such as clicking, scrolling, and typing. The tool was initially powered by GPT-4o but has now transitioned to the more advanced o3 model, promising better accuracy, efficiency, and structured responses.
Early benchmarks highlight notable improvements. On the OSWorld test, which evaluates task completion in browser environments, the o3 model scored 42.9, up from 38.1. The leap is even more pronounced in GAIA assessments, where performance surged from 12.3 to 62.2. User preference studies also favor the new model for its clarity and detail—for instance, when booking restaurants, it provides comprehensive listings with Michelin ratings and seating details in a neatly formatted table.
Safety remains a priority, with the upgraded Operator maintaining strict safeguards. It now confirms 94% of sensitive actions before execution, including a 100% verification rate for financial transactions. High-risk activities, such as accessing email or banking platforms, still require user oversight via Watch Mode or may be blocked entirely. OpenAI has also reduced prompt injection vulnerabilities, though the system remains in “research preview,” indicating ongoing refinements.
For enterprise users, the o3-powered Operator offers tangible workflow benefits. AI engineers can streamline model validation, data teams can automate web scraping with greater reliability, and security professionals gain a safer tool for simulating user behavior during audits. While the Responses API continues using GPT-4o for now, this upgrade underscores OpenAI’s focus on balancing innovation with responsible deployment—a critical consideration as AI agents take on more real-world tasks.
(Source: VentureBeat)
