Openai begins new tools to help enterprises build AI agents

Tuesday, Openai, developers and enterprises AI agents – Independent automated systems that can perform tasks – the company has introduced new tools designed to use its AI models and frames.

API’s new answers are part of API’s new answers that can scan through the company files and walking websites and walk websites and walk websites Openai operator product. Answers are effectively substituted an API Assistants APIThe company planned to sunset in the first half of 2026.

The hype around AI agents, despite the fact that the technological industry is struggling to show people, has grown sharply in recent years, or identifyHow much are “AI agents”. In the latest example of the ugent hype, the Chinese startup butterfly effect this week has gone viral this week For a new AI agent platform called Manus Quickly revealed that many companies have not delivered a lot of promises.

In other words, the shares are high in Openai to get agents.

“It is very easy to demo your agency,” Olivier Godemont, Openai’s API product, speaking to Techcrunch in an interview. “The agency is very difficult to scale and most difficult to use.”

At the beginning of this year, Openai introduced two AI agencies Chatgpt: Operator navigating websites from your behalf and Deep researchResearch compiles research reports for you. Both vehicles suggested a possible idea to achieve the technology, but a little desirable in the “Autonomy” department.

With API answers, Openai wants to sell access to the components of power, which allows developers to build their operator and deep research style agents. OpenTai hopes that developers can create some applications with agent technology that are more autonomous than existing ones today.

Answers using API can be used by the Developers Openai Header (Preview) CHATGPT Search Web search tool: Search GPT-4O and search for GPT-4O mini. The models can view the Internet for answers to the questions, referring to sources for answers.

Openai, GPT-4O search and GPT-4O search claims very true accuracy. For comparison, comparison of models, facts looking for facts, 70% from 90% to 90% to 90% to 88%, GPT-4.5 – Openai, the recently broadcast model is larger, gains a total of 63%.

It is not surprisingly surprising that the traditional AI models of search tools operating AI-power – in theory, GPT-4O search can only search the correct answer. However, the website does not create an unresolved problem of hallucinations. Outside of the actual accuracy, AI search facilities also fight short, navigation requests (eg “Lakers sister” and the latest reports show that Chatgept’s quotes are not always valid.

Answers include a file search program that can quickly scan files in a company’s database to get information. (Openai claims that these files will not develop models. The model creates mouse and keyboard movements that allow developers to automate computer use tasks such as data entry and application work flows.

Enterprises, research previews, local can ask the Cua model, which is widely published in their systems. The consumption version of the CUA available in the operator can only move on the Internet.

Answers to be clear The API will not solve all technical problems plagating AI agents today.

EU-energetic search supplies are more accurate than traditional AI models – a surprising truth can only look for the correct answer – the website does not show AI hallucinations are a solution to a solution. GPT-4O search is still 10% of the wrong questions. Outside their accuracy, AI search tools are also inclined Fight with short, navigation surveys (eg “Today today”) and the latest reports show that Chatgept’s quotes are not always valid.

In a blog post given to TechCrunch, Openai said the Cua model is “not highly valid for automating operating systems” and “careless” mistakes.

However, Openai said they have early iterations of agents and are constantly trying to improve them.

Along with the API, Openai, Openai provides free tools to enter models with internal systems, protecting the activities of the AI ​​Agent for the purposes of the AI ​​agent, the agents called SDK launches an open source set. Agents SDK is a framework for the multifentious orchestra of the company released by the company at the end of last year, the Openai’s swarm.

Godemont, Openai’s gap between the demos and products of the EU agent this year, this year, “Agents will take place,” he said. It’s Sam Altman Sam Altman in January, Echoes Echoes in January: This is the year 2025, AI agents entering the workforce.

Whether 2025 is a “year of AI Agent”, Openai’s latest releases show that the company wants to move from Flashy Agent Demos to effective means.