With its new model, OpenAI promises a less preachy and more relevant ChatGPT.

Depending on the questions asked, ChatGPT will use one of the AI models available to it: Instant for quick answers, Thinking for taking time to reflect, and there is also the possibility of selecting older models. Instant is the most used model, and OpenAI has just refreshed it: goodbye GPT-5.2, welcome GPT-5.3 !

OpenAI revises the ChatGPT version

This model focuses on aspects that users experience daily: tone, relevance, and flow of conversations. GPT-5.3 Instant improves ChatGPT to avoid frustrating responses or simply the absence of responses when the bot should safely reply. It also aims to eliminate the « too cautious or preachy tone » that sometimes characterizes certain interactions, « especially on sensitive topics ».

The « useless refusals » will be significantly reduced, promises OpenAI, as well as the « overly defensive or moralizing preambles » before answering a question. Another interesting change that should make a difference in daily use, ChatGPT should be less insistent, limit « unjustified assumptions » about the user's emotions or intentions, and tone down formulaic phrases that add nothing, such as « Stop. Take a deep breath ».

As with every model update, OpenAI also announces fewer hallucinations… We'll see how it works in practice! And GPT-5.3 Instant should also be a bit more helpful for writing tasks.

Finally, the model is intended to improve the quality of responses after a web search; ChatGPT no longer simply summarizes the results of a search on current events, it will attempt to contextualize them and prioritize the responses deemed most relevant and actionable immediately. All of this is currently in the process of deployment for all users; the Thinking and Pro versions will be available a bit later (they remain with GPT-5.2).

This launch comes at a very complicated time for OpenAI. The company is facing a wave of criticism and uninstallations following the deal signed with the Pentagon to exploit its AI technologies within the context of U.S. military operations.

👉🏻 Follow real-time tech news: add 01net to your Google sources, and subscribe to our WhatsApp channel.

Source : OpenAI