Link: OpenAI claims ChatGPT’s new default model hallucinates way less

OpenAI's GPT-5.5 Instant model demonstrates a reduction in generating false information, making it more reliable compared to its predecessors. The internal evaluation highlighted a 52.5% drop in hallucinations on high-stakes topics such as medicine and law.

The updated model also excels in daily tasks, showing improved responsiveness and factual accuracy. It avoids unnecessary elements like emojis, focusing instead on concise and relevant answers.

GPT-5.5 Instant enhances user interaction by integrating personal data sources like previous chats and Gmail for tailored responses. This feature mirrors efforts by companies like Google with their Gemini system, emphasizing the importance of context in user interactions.

A new 'memory sources' feature will allow users to see and edit the data influencing the chatbot's responses, ensuring greater transparency and control. This feature is currently being introduced to ChatGPT plans on the web and will soon be available on mobile platforms.

Starting Tuesday, OpenAI will implement GPT-5.5 Instant across all ChatGPT user interfaces, with GPT-5.3 remaining available for an additional three months. This phased approach helps users adapt smoothly to the new model.

Enhanced personalization features will initially be available to Plus and Pro users on the web and are expected to extend to mobile and other user tiers soon. This gradual rollout aims to refine user experience across various platforms #

--

Yoooo, this is a quick note on a link that made me go, WTF? Find all past links here.