• Source:JND
Add as a preferred

ChatGPT Agent: The world of AI is continuously evolving, and every big tech company that is into this business is upgrading and evolving their AI models at a faster pace than ever. Currently the buzzword in the industry is AI agents. These are, to put it simply, tools that go beyond your basic chatbot framework and further delve into complex tasks, even multi-step tasks, on your behalf. OpenAIon Thursday unveiled its ChatGPT Agent, a tool that is capable of completing your work or tasks using what the company is calling its own "virtual computer".

The company says that the new agent is capable of automatically navigating through your calendars, generating presentations in an editable format and even running code.

ALSO READ: Dark Blue And Orange iPhone 17 Pro And Pro Max? Leaked Lens Protectors Hint At Striking New Colours!

OpenAI is taking a major leap toward making ChatGPT more than just a chatbot. With the introduction of the ChatGPT agent, the tool now goes beyond answering questions—it can actually take action on your behalf.

This new feature blends capabilities from previous agentic tools like Operator, which could navigate websites, and Deep Research, which could compile insights from multiple sources into a clean, focused report. Users can interact with the agent just by using natural language, the same way they already talk to ChatGPT.

Starting Thursday, the ChatGPT agent will be available for users on Pro, Plus, and Team plans. You can turn it on by selecting “agent mode” from the dropdown menu in the app.

This launch marks OpenAI’s most ambitious push yet to build ChatGPT into a truly task-completing assistant. While many companies—OpenAI, Google, Perplexity, and others—have introduced AI agents with big promises, early versions have often fallen short when faced with more complex jobs. The vision remains exciting, but the path to getting there is still unfolding.

What Is It Capable Of?

This new agent has access to ChatGPT connectors, and this allows users to connect several apps like Gmail and GitHub so that the agent can find out relevant information regarding your prompt. The company claims that the ChatGPT agent can also use APIs to gain access to certain apps.

Let's assume you want to enjoy an English breakfast. OpenAI claims that you can just command ChatGPT Agent to "plan and buy ingredients to make English breakfast for 5" or do some sort of competitor analysis for you. Just type "analyse four competitors and create a presentation". These tasks go beyond just answering your questions; they trigger the AI tool to peruse through the internet and come up with a plan.

How Good Is It?

Well, there always is an ethical concern, but as far as benchmarks and numbers go, OpenAI claims that its new agent offers next-level performance, acing through many benchmarks.

According to OpenAI, the new ChatGPT agent model is showing impressive results on some of the toughest benchmarks out there. It scored 41.6% on Humanity’s Last Exam (pass@1), a notoriously difficult test that spans thousands of questions across over 100 different subjects. That’s nearly twice the performance of OpenAI’s earlier models, o3 and o4-mini.

When it comes to math, the agent also sets a new bar. On FrontierMath—one of the most challenging math tests available—the ChatGPT agent reached a score of 27.4% when equipped with tools like a code-executing terminal. For comparison, the previous best score was just 6.3%, held by o4-mini.

What About Safety and Ethical Usage?

As I wrote above, there are some safety concerns when it comes to these advanced AI tools, and the company claims that it has kept safety in mind before developing this AI agent. OpenAI has previously also warned about the dangers of agentic tools and what the newfound capabilities can do if they get into the wrong hands.

OpenAI has built in a new set of safety measures for its ChatGPT agent, including a real-time monitoring system that keeps an eye on how users interact with the tool. Every prompt entered goes through a classifier to check if it’s related to biology. If it is, the agent’s response gets sent through an additional layer of review to assess whether it could pose a biological threat.

ALSO READ: iQOO Z10R 5G India Launch Set: AMOLED Display, 4K Vlog Camera, And Big Battery Revealed

To further reduce the risk of misuse, OpenAI has also disabled the agent’s memory feature. While memory in other parts of ChatGPT lets it recall details from previous conversations, OpenAI says that in the case of the agent, this could be exploited by attackers to leak sensitive data using prompt injection techniques. The company hasn’t ruled out bringing memory back at a later stage, but for now, it’s staying off.

Despite the hype, the real test for ChatGPT agent will be how well it performs in the real world. So far, most agent technologies have struggled with reliability when handling real tasks. Still, OpenAI believes it’s built a more advanced system that’s ready to take on those challenges.