OpenAI Tests GPT-5 On Human Work: Benchmark Results Show AI Is Able To Equalize Experts

GPT-5 is able to match even beyond the performance of professionals (photo: dock. openai_)

JAKARTA - OpenAI is again making a breakthrough through their latest model, GPT-5. In the latest trial using a benchmark called GDPval, this AI was tested on various real jobs in nine important industries. The results are quite surprising: GPT-5 is able to match even beyond the performance of professionals by 40% of the total testing.

Benchmark GDPval is designed to measure the performance of AI models on tasks normally carried out by humans in the world of work. OpenAI explains that this test includes work from the health, financial, manufacturing, and government sectors. The tasks given are not limited to simulations, but are actually taken from real work practices.

In one test, for example, professionals are asked to compare the reports made by humans with the AI version reports. There are also tests in the investment banking sector, where participants are asked to make a computational analysis in the last distance delivery industry, then the results are aligned with the GPT-5 report.

As a result, GPT-5 has become the OpenAI model with the best performance so far. In 40.6% cases, this AI output is considered equivalent or better than the work of experts in its field. Even so, OpenAI also noted that its competitor, Claude AI of Anthropic, recorded a higher number of 49%. However, according to OpenAI, this is partly because Claude is more skilled at producing visuals and interesting graphs.

So, does this mean that AI will replace humans soon? OpenAI asserts that this has not happened in the near future. According to Dr. Aaron Chatterji, Head of OpenAI Economist, the goal of GDPval is not to prove that AI can fully take over human work. On the other hand, AI is expected to be a supporting tool so that humans can focus more on high-value work.

For example, the task of compiling data-based reports that usually take hours can be completed by GPT-5 in a matter of minutes. That way, workers can allocate their time to a more strategic, creative, or even personal matter.

This GPT-5 achievement marks a transitional phase in the world of work. Instead of seeing it as a threat, OpenAI encourages the use of AI as a work partner that can increase productivity as well as open space for humans to do more meaningful things.