OpenAI officially launches GPT-4.1 model series with many outstanding improvements

OpenAI logo. Photo: AFP/VNA

In the latest announcement, OpenAI introduced the new generation AI model GPT-4.1, and also launched two more reduced versions, GPT-4.1 mini and GPT-4.1 nano, featuring many outstanding improvements in areas such as programming, instruction compliance, and long-term context processing.

According to ChatGPT's developer, this new set of models completely outperforms GPT-4o – OpenAI's previously considered most advanced model – in every aspect, although they are currently only available through the company's API.

Thanks to its enhanced contextual understanding, GPT-4.1 models can process up to 1 million “tokens” – a technical term referring to the units of data that AI can receive and process. Notably, the knowledge base integrated into these models is updated until June 2024.

GPT-4.1's programming performance is reported to increase by 21% compared to GPT-4o and exceed GPT-4.5 by 27%. At the same time, the ability to follow instructions and understand long contexts is significantly improved, making GPT-4.1 models more ideal for supporting AI agents.

“The benchmarks for new models are very high, but OpenAI wants to focus on real -world usefulness,” said X CEO Sam Altman on the social network platform.

OpenAI claims that the operating costs of GPT-4.1 models are significantly lower than those of GPT-4.5. As a result, the company will stop providing the GPT-4.5 preview currently integrated into the API from July 2025, because GPT-4.1 provides comparable or superior performance.

Previously in February 2025, OpenAI released a preview of GPT-4.5 for research purposes to a group of users and programmers, and announced plans to expand access in subsequent phases.

Vietnamese English

Source: https://doanhnghiepvn.vn/cong-nghe/openai-chinh-thuc-trinh-lang-loat-mo-hinh-gpt-4-1-voi-nhieu-cai-tien-vuot-troi/20250417064915808