OpenAI has rolled out an update for its language model GPT-4o, which is used in ChatGPT. Although no detailed release notes were published, the company promises noticeable improvements in the user experience. The update focuses on three core areas: memory management, capabilities in STEM subjects (science, technology, engineering, and mathematics), and conversation management.
The improved memory management should help GPT-4o retain information in context better and respond more relevantly to user requests. In STEM subjects, the model should now provide more precise and well-founded answers. Furthermore, conversation management has been optimized to achieve "productive results." OpenAI aims to make the model more intuitive and effective in various task areas.
Sam Altman, CEO of OpenAI, describes the update as an improvement in both the intelligence and the "personality" of the model. At the same time, he acknowledges that GPT-4o still has weaknesses, especially regarding the understanding of complex relationships. However, Altman assures that OpenAI is working on solutions for these problems.
The lack of detailed release notes is not a new phenomenon. OpenAI has explained in the past that due to methodological challenges, it is difficult to systematically capture and communicate subtle improvements on a large scale. Updating language models is a complex process that resembles a craft rather than an industrial approach.
For developers accessing GPT-4o via the API, the endpoint "chatgpt-4o-latest" continues to point to the most current model. As an alternative for those who prefer a more stable version, OpenAI now offers access to the new GPT-4.1 series. This series is intended to be subject to less frequent updates and thus provide a more reliable basis for applications.
With this update, OpenAI underscores its commitment to the continuous improvement of its language models. The focus on user-friendliness and effectiveness in various application areas shows that the company strives to make the technology accessible and useful for a broad audience. It remains to be seen how the announced improvements will prove themselves in practice and how users perceive the changed interaction with ChatGPT.
Bibliographie: https://the-decoder.com/chatgpt-gets-an-update-openai-promises-a-more-intuitive-gpt-4o/ https://community.openai.com/t/chatgpt-release-notes-2025-march-27-gpt-4o-a-new-update/1153887 https://ground.news/article/chatgpt-gets-an-update-openai-promises-a-more-intuitive-gpt-4o https://www.moneycontrol.com/technology/openai-updates-chatgpt-4-o-model-sam-altman-says-its-personality-and-intelligence-have-improved-article-13005506.html https://opentools.ai/news/openai-ups-its-game-with-gpt-4o-say-hello-to-better-intelligence-and-personality https://help.openai.com/en/articles/6825453-chatgpt-release-notes https://www.techradar.com/computing/artificial-intelligence/chatgpt-4o-just-got-better-although-im-yet-to-notice-a-difference https://www.byteplus.com/en/topic/539908 https://www.msn.com/en-in/money/news/openai-ceo-sam-altman-announces-gpt-4o-got-another-update-in-chatgpt-says-particularly-good-at-coding-and/ar-AA1BOKdk https://opentools.ai/news/openai-ushers-in-a-new-era-with-gpt-4o-launch