OpenAI Operator: Difference between revisions
No edit summary |
|||
| Line 15: | Line 15: | ||
== Performance and limitations == | == Performance and limitations == | ||
In benchmark assessments, Operator achieved notable success, scoring 38.1% on OSWorld benchmarks (OS-level tasks) and 58.1% on WebArena benchmarks (web interactions).<ref name=":0">{{Cite web |last=McFarland |first=Alex |date=2025-01-24 |title=What You Need to Know About OpenAI’s Operator |url=https://www.unite.ai/what-you-need-to-know-about-openais-operator/ |access-date=2025-03-31 |website=Unite.AI |language=en-US}}</ref> However, it | In benchmark assessments, Operator achieved notable success, scoring 38.1% on OSWorld benchmarks (OS-level tasks) and 58.1% on WebArena benchmarks (web interactions).<ref name=":0">{{Cite web |last=McFarland |first=Alex |date=2025-01-24 |title=What You Need to Know About OpenAI’s Operator |url=https://www.unite.ai/what-you-need-to-know-about-openais-operator/ |access-date=2025-03-31 |website=Unite.AI |language=en-US}}</ref> However, it did not reach human-level accuracy and faced limitations with intricate user interfaces and extended workflows.<ref name="OpenAI Blog"/> | ||
== Safety and privacy == | == Safety and privacy == | ||
Revision as of 12:53, 17 December 2025
Template:Short description Template:Infobox software
OpenAI Operator was an AI agent developed by OpenAI, capable of autonomously performing tasks through web browser interactions, including filling forms, placing online orders, scheduling appointments, and other repetitive browser-based tasks.[1][2] It uses OpenAI's advanced models to expand practical automation capabilities for users in daily activities.[3]
Operator was launched on Template:Date.[4] It was released as a limited-access research preview to ChatGPT Pro-tier subscribers in the United States on February 1, 2025, with future plans to broaden availability.[5] Operator was deprecated after the release of ChatGPT agent, and shut down on Template:Date.[6]
Performance and limitations
In benchmark assessments, Operator achieved notable success, scoring 38.1% on OSWorld benchmarks (OS-level tasks) and 58.1% on WebArena benchmarks (web interactions).[7] However, it did not reach human-level accuracy and faced limitations with intricate user interfaces and extended workflows.[3]
Safety and privacy
OpenAI emphasizes privacy and safety measures within Operator, including stringent data protection protocols and built-in safety checks designed to prevent unauthorized sensitive actions or information misuse.[3]
Availability
Initially, Operator was only available to ChatGPT Pro subscribers in the U.S., with plans for broader availability to Plus, Team, and Enterprise users in the future.[7]
References
- ↑ Lin, Belle (2025-01-23). "OpenAI’s ‘Operator’ Agent Can Buy Groceries, File Expense Reports" (in en-US). Wall Street Journal. ISSN 0099-9660. https://www.wsj.com/articles/openais-operator-agent-can-buy-groceries-file-expense-reports-030c30c0.
- ↑ Knight, Will (2025-01-23). "OpenAI’s Operator Lets ChatGPT Use the Web for You" (in en-US). Wired. ISSN 1059-1028. https://www.wired.com/story/openai-sets-chatgpt-loose-on-the-web/.
- ↑ 3.0 3.1 3.2 "Introducing Operator". OpenAI Blog. 2025-02-01. https://openai.com/index/introducing-operator/.
- ↑ Metz, Cade (2025-01-23). "OpenAI Unveils A.I. Agent That Can Use Websites on Its Own" (in en-US). The New York Times. ISSN 0362-4331. https://www.nytimes.com/2025/01/23/technology/openai-operator-launch.html.
- ↑ Kevin Roose (2025-02-01). "OpenAI’s Operator Brings AI to Everyday Web Tasks". The New York Times. https://www.nytimes.com/2025/02/01/technology/openai-operator-tool.html.
- ↑ "Operator - Release Notes" (in en). https://help.openai.com/en/articles/10561834-operator-release-notes.
- ↑ 7.0 7.1 McFarland, Alex (2025-01-24). "What You Need to Know About OpenAI’s Operator" (in en-US). https://www.unite.ai/what-you-need-to-know-about-openais-operator/.