OpenAI releases GPT-4.5 for Pro subscribers
OpenAI made GPT-4.5 available on Thursday (May 28, 2026) for subscribers of the Pro plan, which costs $200 per month. The new model is natively multimodal, capable of processing text, images, and audio simultaneously without relying on separate modules.
According to the company, the unified architecture allows the model to interpret and correlate different types of information in a single pass, eliminating preprocessing steps. Internal data indicates that GPT-4.5 shows a 40% reduction in hallucinations – factually incorrect responses – compared to GPT-4.1. Another highlighted advancement is a 25% improvement in mathematical reasoning tasks.
Access and availability
For now, GPT-4.5 is restricted to the Pro plan, aimed at users who demand high performance. OpenAI has not disclosed timelines for expansion to other plans or the free public.
The launch comes at a time of fierce competition in the generative artificial intelligence sector, with rivals such as Google and Anthropic also investing in multimodal models.
Impact and expectations
Analysts see GPT-4.5 as an important step for applications that require simultaneous understanding of different formats – such as video conferencing, analysis of documents with charts, and descriptions of images in audio. The reduction in hallucinations could expand the model's use in sensitive areas like healthcare and law.
OpenAI states it will continue refining the model and collecting feedback from Pro subscribers before broadening access.