< img src = "/uploads/blogs/13/33/ib-fqt0q6koc_7b271523.jpg" Alt = "Openai launches responses API for business automation"/> ~ ~ ~ ~ ~ ~ ~ < P > Openai has introduced new tools for developers and enterprises that allow you to create automated SI systems capable of performing complex tasks using a new API Responses independent < p > Creating effective SI agents becomes a priority for the technological industry. The developers seek to build autonomous tasks. Many companies declare progress in this area. Not all achievements meet the promises as in the case with the Manus platform.
< p >According to Oli & Rsquo; there is a Godmona, the head of the API products of Openai, the creation of agents' demonstrations is relatively simple. A real challenge is the scale of product and providing regular use. & Amp; nbsp; < strong > precisely to solve these problems aimed at new API Responses. 0 ~/P > > 62 > 62 > 62 > 62 > 62 ~ 62 < h2 > from demonstrations to practical tools
< p > earlier this year Openai introduced two Shi agents in Chatgpt. The operator navigates websites on user. The function of deep research is research reports. Both technologies showed potential but had a standalone restriction.
< p >API Responses gives developers access to the components that provide the work of these agents. This allows you to create your own programs with agent technology. Openai hopes that developers will be able to build more autonomous applications than available now.
< p >< Em > & laquo; scalating the agent is quite difficult, and make people often use it & ndash; It is very difficult & raquo; , & ndash; Notes Godmon. The new API is designed to simplify the process of developing and implementing SI agents. It provides the necessary infrastructure to create practical solutions.
< h2 > key models and functionality
< P > Developers access the models used in the Chatgpt Search tool. It's GPT-4O Search and GPT-4O mini-search. They can search information on the Internet and provide answers with links to sources.
< p > These models show high actual accuracy. In the Simpleqa GPT-4O Search Benchmark, 90%. GPT-4O mini Search showed 88%. For comparison, the standard GPT-4.5 model received only 63% in the same test.
< p >Although search tools are more accurate than traditional models, problems remain. They still give incorrect answers to 10% actual questions. & Amp; nbsp; < strng > Short navigation requests and reliability of quotes also remain problem areas. 0 ~/P > > 62 > 62 ~ 62 < H2 > Additional Opportunities and Tools < P > API Responses also includes a file search utility. It can scan documents in the company databases. Openai emphasizes that he does not teach his models on these files. This feature ensures the confidentiality of corporate data.
< P > Developers also access the Computer-University Agent (Cua). It provides Operator function. The model generates mouse and keyboard actions. This allows you to automate the task of using the computer; Rsquo; Euther.
< p >< Em >& laquo; agents & ndash; This is the most effective application of the AI that will ever occur & raquo; , & ndash; It predicts Godmo. Businesses can launch CUA model locally on their own systems. This provides additional security and control.< H2 > The future of agent technologies
< P > Openai produces Agents SDK with an open source code. This set of tools helps integrate models with internal systems. It also allows you to implement precautions and monitor agents' activities for optimization.
< p > in their statements Openai recognizes current restrictions. Cua & ldquo; not yet high for automation of tasks in operating systems & Rdquo;. The company is working on improving these tools. They are regarded as early iterations of agency technologies.
< p > OpenAi CEOs Sam Altman stated that 2025 would be a year when Shi agents turn into real workforce. Automation.
< p > Regardless of the reconciliation of this forecast, API Responses and Agents SDK reflect the strategic direction of OpenAi development. The company focuses on providing the developers with the necessary tools to build reliable and useful SI agents that can perform real tasks in different areas.