GPT-5.5 Instant: ChatGPT hallucinates significantly less often Apfelpatient

OpenAI is replacing the standard ChatGPT model. GPT-5.5 Instant replaces GPT-5.3 Instant as of today and, according to OpenAI, delivers 52.5 percent fewer fabricated claims in critical topics such as medicine, law, and finance. Another aspect of the behavior has also been removed: excessive emoji use.

OpenAI's pace remains high. Just at the end of April, the company released GPT-5.5 Thinking and Pro, a "new class of intelligence" for demanding knowledge work. Now, the company is following up with GPT-5.5 Instant – the variant that makes all the difference for most ChatGPT users in their daily work. While the Thinking models are relevant for Pro and Enterprise users, Instant is the model behind most ChatGPT conversations. This makes an update particularly significant – also for the Apple ecosystem, since ChatGPT has been integrated into Siri and Image Playground since iOS 18.

What exactly makes GPT-5.5 Instant better?

OpenAI identifies two key areas for improvement. First: significantly fewer hallucinations. In internal tests, GPT-5.5 Instant is said to produce 52.5 percent fewer fabricated statements than its predecessor in critical queries from the fields of medicine, law, and finance. In particularly difficult conversations that users flagged due to factual errors, inaccurate statements are expected to decrease by 37.3 percent. This is remarkable – hallucinations have been the biggest trust issue for AI models since ChatGPT's inception, and the new model is expected to significantly improve precisely in the areas where accuracy is paramount.

Secondly, more concise and natural responses. OpenAI states that the model is "more precise and to the point" without sacrificing substance. The number of follow-up questions has been reduced, as has the tendency towards over-formatting. And one detail explicitly mentioned in the announcement: the "unwarranted" use of emojis has largely disappeared. Anyone who has experienced ChatGPT responses with an unnecessarily high number of smileys, thumbs-up emojis, or lightbulb symbols in recent months will likely notice the difference.

From "less embarrassing" to "factually reliable"

The predecessor, GPT-5.3 Instant, released two months ago, focused on making ChatGPT sound less strenuous and less dramatic. At that time, the goal was to remove exaggerated phrases like "Stop. Take a deep breath" from the default behavior. GPT-5.5 Instant now shifts the focus from tone to substance. This is a logical step: first, the model's sound was cleaned up; now, its factual accuracy is being improved.

It's striking how quickly OpenAI iterates. There are exactly two months between GPT-5.3 Instant at the beginning of March and GPT-5.5 Instant today. In the same period, GPT-5.4 Thinking, GPT-5.5 Thinking, and GPT-5.5 Pro were released. Anyone using ChatGPT as a private user has probably lost track of the version family by now – but that's precisely OpenAI's point: Auto-routing decides in the background which model responds, and the default model simply keeps improving.

What this means for Apple users

For the Apple ecosystem, this update is relevant for two reasons. First, ChatGPT has been integrated as a Siri fallback and into Image Playground since iOS 18. Anyone asking Siri a more complex question or challenging Apple Intelligence beyond a single point will be directed to OpenAI's models in the background – and these are now more reliable.

Secondly, with the Extensions system in iOS 27, Apple is expanding its AI platform for third-party developers, and ChatGPT remains a key player. The more reliably GPT 5.5 Instant responds, the more attractive it becomes to set ChatGPT as the default model for native Apple functions when iOS 27 arrives this fall. Apple itself likely has an interest in this as well: A reliable external AI makes its own voice assistant less vulnerable while Apple's Foundation Models team is still catching up.

Availability and transition

GPT-5.5 Instant is being rolled out to all ChatGPT users starting today – including free users. Paid customers will retain access to GPT-5.3 Instant for an additional three months if they wish to continue using the older model. After that, the previous version will be deactivated. Regular ChatGPT users should notice the change starting today – primarily in the form of shorter replies and fewer random emoticons.

OpenAI's iteration pace remains the real statement

Today's release, in itself, is a solid model update. But in the bigger picture, it reveals something more important: OpenAI is iterating at a pace that is putting the industry under pressure. Halving hallucinations, reducing emojis, reworking tone of voice – all every two weeks. For Apple, Google, and Anthropic, that's the real news of the day. (Image: Shutterstock / Primakov)