OpenAI Previews GPT-5.4 Just 24 Hours After Launching GPT-5.3 Instant

OpenAI is accelerating its artificial intelligence research because it plans to release its next major system update following the recent launch of ChatGPT. The company recently rolled out GPT-5.3 Instant and has already hinted that the next iteration, GPT-5.4, may arrive sooner than many expected.

The announcement of GPT-5.3 Instant focused on refining the everyday user experience inside ChatGPT. The model update is designed to generate quicker responses while producing answers that are more accurate and easier to follow. Engineers also aimed to reduce so-called “hallucinations,” a long-standing issue in generative AI that causes models to provide wrong information with high confidence. 

According to reported benchmarks, GPT-5.3 Instant lowered hallucination rates on web-based queries by about 26.8 percent and improved reliability in knowledge responses by nearly 19.7 percent. The model was developed to enhance factual correctness while building direct communication skills. Earlier versions of the chatbot were sometimes criticized for overly cautious or lengthy replies. The updated system attempts to address those concerns by reducing unnecessary refusals and focusing more clearly on the user’s request.

OpenAI announced its new model to users while it was already working on developing a more powerful system. Company representatives suggested that GPT-5.4 could be released “sooner than people think,” although detailed technical specifications were not immediately disclosed at the time of the teaser announcement.

The latest details about GPT-5.4 show the system functions as a more powerful solution for professional work and complex tasks. The preliminary tests show that the new model has better knowledge-work capabilities because it performs at least at the same level as human experts for 83 percent of the tests, which cover 44 different job positions.

The new model delivers better reasoning abilities and enhanced coding skills, together with software environment and web interface interaction tools. Developers can use these abilities to create AI agents that will perform independent work by completing multi-step operations throughout digital platforms.