OpenAI Upgrades ChatGPT’s Default AI Model to GPT-5.5 Instant, Adds New Capabilities

OpenAI, on Tuesday, announced that it is updating the default artificial intelligence (AI) model in ChatGPT. The default model is available to everyone when they first open the website or the app, including those on the free tier. So far, this experience has been powered by the GPT-5.3 Instant, but now, the San Francisco-based AI giant is replacing it with the GPT-5.5 Instant. Some of the key improvements include more personalised responses, higher intelligence in image analysis and answering science and math questions, and a natural conversational tone.

GPT-5.5 Instant Is Now Rolling Out

In a post, the AI giant announced that the default ChatGPT model will now be GPT-5.5 Instant. It is part of the GPT-5.5 family, which was first released last month. The Instant variant is designed to be more responsive and conversational compared to the other models in the series. “Because Instant is the daily driver for hundreds of millions of people, small improvements make a big difference,” the post said.

One of the most notable improvements is the large language model's (LLM's) reduced inaccuracy and higher intelligence when it comes to image analysis, and science, technology, engineering and mathematics (STEM)-related questions. It is also said to be producing 52.5 percent fewer hallucinated claims compared to its predecessor.

Another area of improvement is conversational tone. Instead of being needlessly chatty and verbose, the newer model is said to be “more to-the-point without losing substance.” OpenAI says the model keeps its warmth, but can now deliver more utility using the same information. Overformatting responses will also be fewer with this version of the model. “It also asks fewer unnecessary follow-up questions and avoids things that can make responses feel cluttered, like gratuitous emojis,” OpenAI said.

Personalisation is also being improved with GPT-5.5 Instant. OpenAI says the model is better at drawing context from past chats, files, and Gmail, provided the user has connected the chatbot with the email client. The model also decides on its own when a response will be benefited by personalisation, and when it might be unnecessary. One of the highlighted use cases is when a user asks for suggestions or recommendations.