OpenAI Releases GPT-5.3 Instant Model to Fix ChatGPT Hallucinations
In Focus
- ChatGPT-5.3 Instant model replaces GPT-5.2 Instant model
- Users complained about cringe responses generated by the GPT-5.2 Instant model
- Some users unsubscribed from ChatGPT due to this problem
OpenAI has released the ChatGPT-5.3 Instant model. According to TechCrunch, the new AI model is designed to reduce cringe responses and disclaimers. The AI developer said that the GPT-5.3 update enhances user experience through improvements to relevance, tone, and conversational flow.
“GPT‑5.3 Instant delivers more accurate answers, richer and better-contextualized results when searching the web, and reduces unnecessary dead ends, caveats, and overly declarative phrasing that can interrupt the flow of conversation,” OpenAI stated.
New Update Addresses User Concerns
The latest GPT-5.3 Instant update reduces hallucinations in OpenAI’s leading chatbot, which has been a major concern for users. The AI developer acknowledged that users had provided feedback on the matter, noting that users can now expect less cringe in ChatGPT responses.
“We heard feedback that GPT‑5.2 Instant would sometimes respond in ways that feel overly cautious or preachy, particularly around sensitive topics. GPT‑5.3 Instant significantly tones down overly defensive or moralizing preambles before answering the question,” OpenAI added.
Many users have been frustrated by the tone of ChatGPT’s 5.2 model. They described it as irritating and hard to tolerate. OpenAI released the GPT-5.3 model update weeks after rolling out ads on ChatGPT for all free users.
Users Unsubscribed from ChatGPT
Complaints spread across social media, with some users saying they had canceled their ChatGPT subscriptions. Some users also claimed that ChatGPT responses assumed they were anxious or overwhelmed, even when they were simply asking for information.
Instead of generating straightforward answers, the chatbot would suggest actions like taking a deep breath or reassuring them unnecessarily. For some users, this tone was condescending; others said it made them feel talked down to. Prior to the ChatGPT conversational flow update, the issue had become a topic of discussion on Reddit.
GPT 5.3 Acknowledges Challenges
OpenAI demonstrated the difference in query responses between GPT-5.2 Instant and GPT-5.3 Instant models. In the example, GPT-5.2 Instant starts the response with a “First of all, you’re not broken,” phrase, which frequently upset users. On the contrary, GPT-5.2 acknowledges the challenge of a situation without reassuring the user.
This shift is critical for OpenAI due to the legal action that the company is facing. Currently, the AI developer is dealing with lawsuits over claims that ChatGPT contributed to harmful mental health outcomes, including suicide.
But there’s a fine line between being empathetic and simply giving clear, direct answers. When people look up information, they expect facts, not emotional check-ins or assumptions about how they’re feeling.
